<?xml version="1.0" encoding="utf-8"?>
<rss version="2.0" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/">
    <channel>
        <title>karmada Blog</title>
        <link>https://karmada.io/zh/blog</link>
        <description>karmada Blog</description>
        <lastBuildDate>Sat, 06 Jun 2026 00:00:00 GMT</lastBuildDate>
        <docs>https://validator.w3.org/feed/docs/rss2.html</docs>
        <generator>https://github.com/jpmonette/feed</generator>
        <language>zh</language>
        <item>
            <title><![CDATA[Karmada v1.18 版本发布！新增混合云溢出式调度能力]]></title>
            <link>https://karmada.io/zh/blog/2026/06/06/karmada-v1.18/karmada-v1.18</link>
            <guid>https://karmada.io/zh/blog/2026/06/06/karmada-v1.18/karmada-v1.18</guid>
            <pubDate>Sat, 06 Jun 2026 00:00:00 GMT</pubDate>
            <description><![CDATA[Karmada 是开放的多云多集群容器编排引擎，旨在帮助用户在多云环境下部署和运维业务应用。凭借兼容 Kubernetes 原生 API 的能力，Karmada 可以平滑迁移单集群工作负载，并且仍可保持与 Kubernetes 周边生态工具链协同。]]></description>
            <content:encoded><![CDATA[<p>Karmada 是开放的多云多集群容器编排引擎，旨在帮助用户在多云环境下部署和运维业务应用。凭借兼容 Kubernetes 原生 API 的能力，Karmada 可以平滑迁移单集群工作负载，并且仍可保持与 Kubernetes 周边生态工具链协同。</p>
<p><a href="https://github.com/karmada-io/karmada/releases/tag/v1.18.0" target="_blank" rel="noopener noreferrer" class="">Karmada v1.18</a> 版本现已发布，本版本包含下列新增特性：</p>
<ul>
<li class="">支持混合云场景下的溢出式集群亲和调度</li>
<li class="">引入调度超分保护机制</li>
</ul>
<p>这些更新进一步增强了 Karmada 在混合云和大规模调度场景中的可用性与灵活性。我们鼓励您升级到 v1.18.0，体验这些新能力带来的价值。</p>
<h1>新特性概览</h1>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="支持混合云场景下的溢出式集群亲和调度">支持混合云场景下的溢出式集群亲和调度<a href="https://karmada.io/zh/blog/2026/06/06/karmada-v1.18/karmada-v1.18#%E6%94%AF%E6%8C%81%E6%B7%B7%E5%90%88%E4%BA%91%E5%9C%BA%E6%99%AF%E4%B8%8B%E7%9A%84%E6%BA%A2%E5%87%BA%E5%BC%8F%E9%9B%86%E7%BE%A4%E4%BA%B2%E5%92%8C%E8%B0%83%E5%BA%A6" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p>在混合云环境中，企业通常会将本地数据中心作为主要资源池，把公有云作为峰值流量到来时的弹性补充资源。此前，Karmada 的 <code>ClusterAffinities</code> 已经支持声明多个候选集群组，但这些集群组在一次调度过程中是互斥的：调度器最终只会选择其中一个集群组，无法在首选资源池容量不足时，将剩余副本继续扩展到补充资源池。</p>
<p>Karmada v1.18 引入了 <strong>Overflow Cluster Affinities</strong> 能力，在 <code>ClusterAffinityTerm</code> 中新增 <code>overflowAffinities</code> 字段，用于声明按优先级排列的补充集群组。调度器会优先填满主集群组；当主集群组资源耗尽后，再按声明顺序逐级向补充集群组溢出副本，从而实现“优先使用 IDC，容量不足时再溢出到云”的混合云调度模式。</p>
<p>这项能力主要带来以下价值：</p>
<ul>
<li class=""><strong>渐进式溢出</strong>：用户可以在同一个 <code>ClusterAffinityTerm</code> 中声明多个补充集群组，调度器会按照顺序逐级扩展调度范围。</li>
<li class=""><strong>反向收缩</strong>：在缩容场景下，副本会优先从补充集群组回收，从而尽量保留主资源池中的稳定负载。</li>
<li class=""><strong>成本优化</strong>：让基线工作负载稳定运行在成本更优的本地集群上，只在确有需要时才使用公有云资源承接突发流量。</li>
</ul>
<p>这一能力特别适用于如下场景：</p>
<ul>
<li class=""><strong>弹性 GPU 调度</strong>：本地 IDC GPU 集群承载日常推理流量，云上 GPU 集群仅用于承接峰值负载。</li>
<li class=""><strong>成本控制</strong>：将常态业务留在本地基础设施，借助公有云弹性处理流量波峰。</li>
<li class=""><strong>容量规划</strong>：用清晰的优先级关系定义资源池使用顺序，减少人工干预。</li>
</ul>
<p>效果图如下：</p>
<p><img decoding="async" loading="lazy" alt="overflow-affinities-scheduling" src="https://karmada.io/zh/assets/images/overflow-affinities-scheduling-d94c8559d6c86046a48e35772abe9170.png" width="892" height="430" class="img_ev3q"></p>
<p>下面是一个精简示例，优先将工作负载调度到 IDC GPU 集群，不足时再溢出到云上 GPU 集群：</p>
<div class="language-yaml codeBlockContainer_Ckt0 theme-code-block" style="--prism-color:#393A34;--prism-background-color:#f6f8fa"><div class="codeBlockContent_QJqH"><pre tabindex="0" class="prism-code language-yaml codeBlock_bY9V thin-scrollbar" style="color:#393A34;background-color:#f6f8fa"><code class="codeBlockLines_e6Vv"><span class="token-line" style="color:#393A34"><span class="token key atrule" style="color:#00a4db">apiVersion</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> policy.karmada.io/v1alpha1</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">kind</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> PropagationPolicy</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">metadata</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">name</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> gpu</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">inference</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">overflow</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">spec</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">resourceSelectors</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">apiVersion</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> apps/v1</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">kind</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> Deployment</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">placement</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">clusterAffinities</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">affinityName</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> idc</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">gpu</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">        </span><span class="token key atrule" style="color:#00a4db">clusterNames</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">          </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> idc</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">gpu</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">cluster1</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">        </span><span class="token key atrule" style="color:#00a4db">overflowAffinities</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">          </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">affinityName</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> cloud</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">gpu</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">            </span><span class="token key atrule" style="color:#00a4db">clusterNames</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">              </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> cloud</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">gpu</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">cluster1</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">              </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> cloud</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">gpu</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">cluster2</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">replicaScheduling</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">replicaSchedulingType</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> Divided</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">replicaDivisionPreference</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> Weighted</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">weightPreference</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">        </span><span class="token key atrule" style="color:#00a4db">dynamicWeight</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> AvailableReplicas</span><br></span></code></pre></div></div>
<p>更多有关此功能的资料请参考：<a href="https://karmada.io/docs/userguide/scheduling/propagation-policy#how-overflowaffinities-works" target="_blank" rel="noopener noreferrer" class="">官方特性文档</a> 和 <a href="https://github.com/karmada-io/karmada/blob/master/docs/proposals/scheduling/multi-scheduling-group/overflow-affinities/README.md" target="_blank" rel="noopener noreferrer" class="">特性提案</a>。</p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="引入调度超分保护机制">引入调度超分保护机制<a href="https://karmada.io/zh/blog/2026/06/06/karmada-v1.18/karmada-v1.18#%E5%BC%95%E5%85%A5%E8%B0%83%E5%BA%A6%E8%B6%85%E5%88%86%E4%BF%9D%E6%8A%A4%E6%9C%BA%E5%88%B6" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p>在大规模多集群环境中，调度器虽然按顺序处理调度请求，但在决定某个集群是否还有可用容量时，需要依赖 <code>karmada-scheduler-estimator</code> 提供的集群容量估算结果。问题在于：调度器刚刚完成一次副本分配后，新的工作负载真正下发到成员集群、创建 Pod 并绑定到节点之间，存在天然的时间窗口。</p>
<p>如果在这个窗口内又到来新的调度请求，<code>scheduler-estimator</code> 看到的仍然可能是“旧的容量快照”，从而导致多个连续调度决策重复使用同一份尚未更新的空闲资源，造成资源超卖。最终结果就是：调度阶段看起来成功，但工作负载落地后可能因为资源不足长期 Pending。</p>
<p>为了解决这一问题，Karmada v1.18 引入了 <strong>Scheduling Overcommit Protection（调度超分保护）</strong>。该特性在调度器和估算器之间引入“assume and deduct（先假定、再扣减）”机制：</p>
<ul>
<li class="">调度器在完成一次调度决策后，会先假定这部分资源已经被占用</li>
<li class="">估算器在后续容量计算时，会把这部分“假定占用”的资源一起考虑进去</li>
<li class="">从而避免后续调度在真实状态尚未更新前重复消耗同一批资源</li>
</ul>
<p>这一特性尤其适用于高吞吐量调度场景，可以显著降低由于状态延迟带来的资源超卖问题，让副本分配结果更贴近成员集群的真实承载能力。</p>
<p>当前该特性是 Alpha 特性，默认关闭，并且需要同时在 <code>karmada-scheduler</code> 和 <code>karmada-scheduler-estimator</code> 上启用：</p>
<div class="language-shell codeBlockContainer_Ckt0 theme-code-block" style="--prism-color:#393A34;--prism-background-color:#f6f8fa"><div class="codeBlockContent_QJqH"><pre tabindex="0" class="prism-code language-shell codeBlock_bY9V thin-scrollbar" style="color:#393A34;background-color:#f6f8fa"><code class="codeBlockLines_e6Vv"><span class="token-line" style="color:#393A34"><span class="token plain">--feature-gates</span><span class="token operator" style="color:#393A34">=</span><span class="token plain">SchedulingOvercommitProtection</span><span class="token operator" style="color:#393A34">=</span><span class="token plain">true</span><br></span></code></pre></div></div>
<p>需要注意的是，该特性只对 <code>ReplicaSchedulingType: Divided</code> 且采用容量感知分配策略的场景生效，例如 <code>Aggregated</code> 或 <code>DynamicWeight</code>。对于 <code>Duplicated</code> 模式以及完全静态权重分配的场景，不会产生影响。</p>
<p>更多有关此功能的资料请参考：<a href="https://karmada.io/zh/docs/userguide/scheduling/scheduling-overcommit-protection/" target="_blank" rel="noopener noreferrer" class="">调度超分保护</a>。</p>
<h1>致谢贡献者</h1>
<p>Karmada v1.18 版本包含了来自 31 位贡献者的 144 次代码提交，在此对各位贡献者表示由衷的感谢：</p>
<table><thead><tr><th style="text-align:left">^-^</th><th style="text-align:left">^-^</th><th style="text-align:left">^-^</th></tr></thead><tbody><tr><td style="text-align:left">@abhicodes11</td><td style="text-align:left">@Ady0333</td><td style="text-align:left">@dahuo98</td></tr><tr><td style="text-align:left">@Denyme24</td><td style="text-align:left">@faucct</td><td style="text-align:left">@FAUST-BENCHOU</td></tr><tr><td style="text-align:left">@gjbravi</td><td style="text-align:left">@hl8086</td><td style="text-align:left">@jabellard</td></tr><tr><td style="text-align:left">@Krishiv-Mahajan</td><td style="text-align:left">@LivingCcj</td><td style="text-align:left">@manmathbh</td></tr><tr><td style="text-align:left">@mszacillo</td><td style="text-align:left">@nXtCyberNet</td><td style="text-align:left">@Park-Jiyeonn</td></tr><tr><td style="text-align:left">@qiuming520</td><td style="text-align:left">@RainbowMango</td><td style="text-align:left">@seanlaii</td></tr><tr><td style="text-align:left">@SujoyDutta</td><td style="text-align:left">@SunsetB612</td><td style="text-align:left">@Tej-Katika</td></tr><tr><td style="text-align:left">@tessapham</td><td style="text-align:left">@vanshiz</td><td style="text-align:left">@vgt-rangehrn</td></tr><tr><td style="text-align:left">@vie-serendipity</td><td style="text-align:left">@Vinayak9769</td><td style="text-align:left">@Viscous106</td></tr><tr><td style="text-align:left">@warjiang</td><td style="text-align:left">@whitewindmills</td><td style="text-align:left">@XiShanYongYe-Chang</td></tr><tr><td style="text-align:left">@zhzhuang-zju</td><td style="text-align:left"></td><td style="text-align:left"></td></tr></tbody></table>
<p><img decoding="async" loading="lazy" alt="贡献者" src="https://karmada.io/zh/assets/images/contributors-e90e7caebfb331363959059670c14949.png" width="597" height="695" class="img_ev3q"></p>
<h1>参考资料</h1>
<p>[1] Karmada v1.18 发布页: <a href="https://github.com/karmada-io/karmada/releases/tag/v1.18.0" target="_blank" rel="noopener noreferrer" class="">https://github.com/karmada-io/karmada/releases/tag/v1.18.0</a></p>
<p>[2] Karmada v1.18 发行说明: <a href="https://github.com/karmada-io/karmada/blob/master/docs/CHANGELOG/CHANGELOG-1.18.md" target="_blank" rel="noopener noreferrer" class="">https://github.com/karmada-io/karmada/blob/master/docs/CHANGELOG/CHANGELOG-1.18.md</a></p>
<p>[3] 溢出式集群亲和调度特性文档: <a href="https://karmada.io/zh/docs/userguide/scheduling/propagation-policy/#overflowaffinities-%E5%B7%A5%E4%BD%9C%E5%8E%9F%E7%90%86" target="_blank" rel="noopener noreferrer" class="">https://karmada.io/zh/docs/userguide/scheduling/propagation-policy/#overflowaffinities-%E5%B7%A5%E4%BD%9C%E5%8E%9F%E7%90%86</a></p>
<p>[4] 溢出式集群亲和调度提案: <a href="https://github.com/karmada-io/karmada/blob/master/docs/proposals/scheduling/multi-scheduling-group/overflow-affinities/README.md" target="_blank" rel="noopener noreferrer" class="">https://github.com/karmada-io/karmada/blob/master/docs/proposals/scheduling/multi-scheduling-group/overflow-affinities/README.md</a></p>
<p>[5] 调度超分保护特性文档: <a href="https://karmada.io/zh/docs/userguide/scheduling/scheduling-overcommit-protection" target="_blank" rel="noopener noreferrer" class="">https://karmada.io/zh/docs/userguide/scheduling/scheduling-overcommit-protection</a></p>
<p>[6] 调度超分保护提案: <a href="https://github.com/karmada-io/karmada/blob/master/docs/proposals/scheduling/estimator-reservation/README.md" target="_blank" rel="noopener noreferrer" class="">https://github.com/karmada-io/karmada/blob/master/docs/proposals/scheduling/estimator-reservation/README.md</a></p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[欢迎 Wellhub 加入 Karmada 用户组]]></title>
            <link>https://karmada.io/zh/blog/2026/05/12/wellhub-officially-joined-the-karmada-adopter-group/wellhub</link>
            <guid>https://karmada.io/zh/blog/2026/05/12/wellhub-officially-joined-the-karmada-adopter-group/wellhub</guid>
            <pubDate>Tue, 12 May 2026 00:00:00 GMT</pubDate>
            <description><![CDATA[Karmada 非常高兴地宣布 Wellhub 正式加入 Karmada 用户组，成为社区的重要成员。作为云原生计算基金会（CNCF）旗下的项目，Karmada 致力于为用户提供强大的多集群管理和调度能力，帮助企业在复杂的分布式环境中实现高效的应用部署和管理。Wellhub 的加入将进一步加强 Karmada 社区，为项目的持续创新带来新的活力，标志着我们社区发展和 Karmada 在多样化生产环境中采用的又一个重要里程碑。]]></description>
            <content:encoded><![CDATA[<p>Karmada 非常高兴地宣布 <strong><a href="https://wellhub.com/" target="_blank" rel="noopener noreferrer" class="">Wellhub</a></strong> 正式加入 Karmada 用户组，成为社区的重要成员。作为云原生计算基金会（CNCF）旗下的项目，Karmada 致力于为用户提供强大的多集群管理和调度能力，帮助企业在复杂的分布式环境中实现高效的应用部署和管理。<strong>Wellhub</strong> 的加入将进一步加强 Karmada 社区，为项目的持续创新带来新的活力，标志着我们社区发展和 Karmada 在多样化生产环境中采用的又一个重要里程碑。</p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="关于-wellhub">关于 Wellhub<a href="https://karmada.io/zh/blog/2026/05/12/wellhub-officially-joined-the-karmada-adopter-group/wellhub#%E5%85%B3%E4%BA%8E-wellhub" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p><img decoding="async" loading="lazy" alt="wellhub logo" src="https://karmada.io/zh/assets/images/wellhub-e6db3ccc6fcaf0321f0626f2c6c229ee.png" width="3840" height="2160" class="img_ev3q"></p>
<p>Wellhub 秉持一大使命：助力每家企业转型为健康型企业，让员工每日关注自身身心健康状态。为此，平台推出高性价比、一站式综合服务方案，全面守护员工的身体、心理、营养与情绪健康。依托 Wellhub，员工无论居家还是在职场，都能主动建立并长期坚持健康的日常生活作息。</p>
<p>目前，全球 11 个国家超 19000 家企业已选用 Wellhub，为数百万员工提供行业顶尖的企业健康福利项目。该项目经实践验证，能大幅提升员工参与度与投入度。</p>
<p>Wellhub 可使参与健康管理的员工数量实现翻倍增长。员工的广泛参与，能让企业人员流失率降低 40%，同时为企业节省最高 35% 的医疗相关开支。</p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="关于-karmada-用户组">关于 Karmada 用户组<a href="https://karmada.io/zh/blog/2026/05/12/wellhub-officially-joined-the-karmada-adopter-group/wellhub#%E5%85%B3%E4%BA%8E-karmada-%E7%94%A8%E6%88%B7%E7%BB%84" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p><strong>Karmada 用户组</strong>是一个由在其环境中成功采用 Karmada 的组织和用户组成的社区。成员分享他们的经验、最佳实践和反馈，以帮助改进 Karmada 并培育一个充满活力的协作生态系统。</p>
<p>成为 Karmada 用户组成员具有以下几个优势：</p>
<ul>
<li class=""><strong>社区认可</strong>：作为云原生多集群管理领域的领导者来展示您的组织，在 CNCF 和 Karmada 社区中获得知名度。</li>
<li class=""><strong>促进协作</strong>：与其他采用者建立联系，分享最佳实践，并在实际用例和解决方案上进行协作。</li>
<li class=""><strong>保持信息同步</strong>：及时接收重要更新通知，包括关键功能、错误修复和安全建议。</li>
<li class=""><strong>活动参与</strong>：受邀参与 Karmada 相关活动，包括 KubeCon + CloudNativeCon、网络研讨会和聚会。</li>
<li class=""><strong>职位发布</strong>：有机会在 Karmada 社区支持的职位公告板上发布与 Karmada 相关的职位空缺（目前暂不可用）。</li>
<li class=""><strong>扩展商业机会</strong>：与 Karmada 生态系统的其他成员建立潜在的商业联系和合作。</li>
</ul>
<p>在 <a href="https://github.com/karmada-io/community/tree/main/adopter-group" target="_blank" rel="noopener noreferrer" class="">GitHub 社区仓库</a> 中了解更多关于 Karmada 用户组的信息，并在 <a href="https://karmada.io/zh/adopters" target="_blank" rel="noopener noreferrer" class="">karmada.io/adopters</a> 查看完整的<code>公开</code>的采用者列表。</p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="加入-karmada-用户组">加入 Karmada 用户组<a href="https://karmada.io/zh/blog/2026/05/12/wellhub-officially-joined-the-karmada-adopter-group/wellhub#%E5%8A%A0%E5%85%A5-karmada-%E7%94%A8%E6%88%B7%E7%BB%84" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p><strong>Karmada 用户组</strong>对当前正在<strong>生产环境</strong>中使用 Karmada 的<strong>最终用户</strong>和<strong>供应商</strong>开放。这包括：</p>
<ul>
<li class=""><strong>最终用户</strong>：在其生产环境中运行 Karmada 的组织。</li>
<li class=""><strong>供应商</strong>：提供基于 Karmada 的产品或服务，并有客户在生产环境中使用这些产品或服务的公司。</li>
</ul>
<p>您是否在生产环境中使用 Karmada 并有兴趣加入 Karmada 用户组？</p>
<p>加入很简单——只需填写 <a href="https://github.com/karmada-io/karmada/issues/4540" target="_blank" rel="noopener noreferrer" class="">Karmada 用户组申请表单</a> 并提供您组织的信息。在您的申请经过审核和批准后，您的组织将被添加到 Karmada 用户组。</p>
<p>更多详情，请查看社区仓库中的 <a href="https://github.com/karmada-io/community/tree/main/adopter-group#how-to-join" target="_blank" rel="noopener noreferrer" class="">如何加入</a> 部分。</p>
<p>让我们一起发展 Karmada 社区和更广泛的多集群生态系统！</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[欢迎 GMI Cloud 加入 Karmada 用户组]]></title>
            <link>https://karmada.io/zh/blog/2026/04/24/gmi-cloud-officially-joined-the-karmada-adopter-group/gmi-cloud</link>
            <guid>https://karmada.io/zh/blog/2026/04/24/gmi-cloud-officially-joined-the-karmada-adopter-group/gmi-cloud</guid>
            <pubDate>Fri, 24 Apr 2026 00:00:00 GMT</pubDate>
            <description><![CDATA[Karmada 非常高兴地宣布 GMI Cloud 正式加入 Karmada 用户组，成为社区的重要成员。作为云原生计算基金会（CNCF）旗下的项目，Karmada 致力于为用户提供强大的多集群管理和调度能力，帮助企业在复杂的分布式环境中实现高效的应用部署和管理。GMI Cloud 的加入将进一步加强 Karmada 社区，为项目的持续创新带来新的活力，标志着我们社区发展和 Karmada 在多样化生产环境中采用的又一个重要里程碑。]]></description>
            <content:encoded><![CDATA[<p>Karmada 非常高兴地宣布 <strong><a href="https://www.gmicloud.ai/" target="_blank" rel="noopener noreferrer" class="">GMI Cloud</a></strong> 正式加入 Karmada 用户组，成为社区的重要成员。作为云原生计算基金会（CNCF）旗下的项目，Karmada 致力于为用户提供强大的多集群管理和调度能力，帮助企业在复杂的分布式环境中实现高效的应用部署和管理。<strong>GMI Cloud</strong> 的加入将进一步加强 Karmada 社区，为项目的持续创新带来新的活力，标志着我们社区发展和 Karmada 在多样化生产环境中采用的又一个重要里程碑。</p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="关于-gmi-cloud">关于 GMI Cloud<a href="https://karmada.io/zh/blog/2026/04/24/gmi-cloud-officially-joined-the-karmada-adopter-group/gmi-cloud#%E5%85%B3%E4%BA%8E-gmi-cloud" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p><img decoding="async" loading="lazy" alt="member logo" src="data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAOIAAABiCAIAAADCwphYAAAQAElEQVR4AeydCdhVUxfHXUIpiqRBg4rSpDRnqAhFpZQmSZMeTaTeZuL7aB6pRHOKNFBIoZTPGA1SURJNSoM0kRC+7/c5j/sc5+617z7nnvt2X+37nM6zz9prr73OOv+z9tpr3ft25hn2Yy2Q8hawME35R2QVPOMMC1OLggxgAQvTDPCQrIoWphYDGcACFqYZ4CFZFdUwtXaxFkgpC1iYptTjsMqoLWBhqraLpaaUBSxMU+pxWGXUFrAwVdvFUlPKAhamKfU4rDJqC/iBqVqCpVoLJN0CFqZJN7GdIHELWJgmbkMrIekWsDBNuontBIlbwMI0cRtaCUm3gIVp0k1sJ0jcAonDNHEdrARrgTgWsDCNYyDbnQoWsDBNhadgdYhjAQvTOAay3algAQvTVHgKVoc4FrAwjWMg250KFkgWTFPh3qwO/xgLnEqYlilTplOnTnPmzNmxY8d/XZ99+/Zt3Lhx0aJF9957b/bs2f8xtrY3EtgCpwCmhQoVGjJkyO7duzds2DBhwoRmzZpBcd9A7ty5S5cuXa9evcmTJx85cuTVV1+98cYb3Qy2fbpZIF1heuGFF44ePRrf2bdv30svvdTQ1vXr11++fDn+9e677zYcYtn+YRZIP5i2b99+27Zt3bt3D2ZB/OusWbM++uijatWqBZOQ7FGFCxe+7bbbevbsOWXKFN6r//z1eemll8aNG9evX7/WrVvXqlWLdzXZmvzz5KcHTHPkyLFgwQIeHo0ELVilSpUPP/zw4YcfTlBOKMMzZ8588803jxw5kpfnp59+4iVcvHjxiBEjeCGJUmr89WnUqFHXrl0HDx48Y8aMt95669ChQ3DOnTs3LS2tUqVKfjXJmzfvv2I+xPG8FKNGjfL01K1b1698Nz8+xSOQjQQTcSMeOrfsHhjb7tWrl2eIc1m9evVY5lhK0mFapEiRTz/99I477vj/3CH9e/zxx1944YVMmTKFJM+3GDz61KlTDx48uHTpUtDGy5MlSxZzKfjdpk2bgu9Vq1YBcdpnnXWW4XBg+mjMhziel6JHjx6enj59+hiKjWXLli0bEZpHIBsJJmJZ8NDZ7MZKcFOAqWeIc4k0N5vUTi5M8RZr1qzx7JAkVXzRmzdvzsLqa0gozNdeey3unKNdu3ZZs2ZNXCYQx7Nu374d13XeeeclLtAtAfubvwDugbSvv/56zsk+SPCYTJFEmBYsWBBnk7xQjPXi6aefNrnJUHhy5sxJlPn+++/jSkMR6BZSoEABXNeWLVvuvPNONz3BNmFJxYoVgwnhhQw20NeoSCRiwp9EmC5cuDBuMEpaatCgQVWrVsUzRf76lC9f/pFHHmH48ePH9ffQsWNH3KqeJ5ReYtBNmzYRZYYiTRKSP3/++fPn824TKUk8f/zxh9SlpAdGW+CBSjUktU+xN2VvC9qUGjtEsqEE+HgR9kMff/wxWxCHznndunVEn2ACf/zkk09C0RzTpk3Dz2kYEu8aOHAg0LnkkksSF2UigVfi66+/zpUrl5KZd1lJl4jB0MYs+A5JZgA6AgOMig5JijdlWWRvG53D0zhx4kSrVq0aNGiwZMkST5fnkk3xgw8+iDS2xp4u55KQjqzq999/71yGfiZYRMmHHnoodMlxBZ577rlKHr/P23CP4pmrQoUKBAweYiKXftX2zJUUmFI98kwTvTx27Bix+XPPPRelnHFGnCZ7YcqqxAAePlI/pUqVSt5GCidNJHrrrbd65tVcfvLJJzNnzhwwYEC3bt3atm3buHFjEqWcu3Tpgkt+7bXXdu3apRmejC7uQhNCSDNKPvj333+XhiSVHj5MW7RoAXqUShNrklBcu3atsldDJCQgBmAH4/CABl733r1745gdSjLO5AivvvrquJL3799PiEKEg8NAK5I1IHLs2LEkF8kWr1ixgjM1YbCL4yfpkS9fPrIExDxxJSsZNMGcFP9JmFPKd4jSkAMHDjgMfs8atU1EhQ/T++67TzkxRmzYsGEAjEal3XXXXW+//TborFy5MkiN0pPR6N+//0033aSXTAzNBi5Pnjxs+GjrmaO9e/funT59OjEPwW7fvn39+ldehqgoT2Pr1q0einN5zTXXOA3zc+gwldSW6B5VQ4Zp0aJFpWCITRU1GM/0vi5//fVXnDFrfbKXHtKNOEiNbjgV6i74TlKeGjZ913fffTds2LArrriCAgFRuIfZ8Pm5R61fv959GW1LmIsyeBqOy/cQnUtu3Gmk8zlkmDZp0kR5AzwGqg7KrhQkzps378wzRcsQfV5++eVkGBJcyJwb590jY8rrTWDgUJyzJFyiM0qCKTEYJSUYDA8J1qyHgWEqqS3RPaqKD8PDZ3iJt1Nyzpo16+jRo8quVCNS97vsssskrcaMGUP0+cMPP0gMwehHjhxhm0VOI66VNF5WgimvnIQ8pbYS88aNG3/77TflkLhESW2J7hEYMkzZxXsmcC6BqdMwP58qTmJfaerBgwdTN5d6E6eT0wCp+/bt04jSuB82c99++61yrIQ8JbMUy1IiVvKbECW1JbpHZpgwLVasmDLZdvLkyUR2Th6Nk3pZp04dgkXlFNQg0iGBunnzZorAZIIlNyPRHZ3fe+89p+E5m8OUfO1VV13lGe5cfvDBB04jwFlSW6J7pggTplLhhMXCM2vKXnbo0EHSjbVe6gqXzoadQtTPP/8cQKzk8HCQhoDgJSFIUE6dCEyVAs2J6QFTSn/mCmk4CxcuTOBI7WDlypXffPMN68Xhw4d5qCThhw8fHkpxT0rmk17YsmWLRrdwu9atW0ceQCmTu1bSHaKEJFY5kxwwQiS/i4PfsWMHDMEOSW2J7pklNJiSxKH04pHuXCa44ciSJQul19WrV1MyBaMgFUTmz58f4Tly5GDTjWV79eoFdvfs2TNo0KC8efPSFeC44YYbmEs5MD2/iqVUIErUP1fSyZIbxkpRIZoGflfZS8ZaSTck6tWOK+TMuBxxGUqUKEGtZdWqVTVr1lQyUxlX0k2I1PR37txJztXkC2kUeEjLs40YNWqUiXAPjzQFgfX27ds9zKfqUr92gwZeV6VuEv7czAiXNsEfJBCYMgWSOQc+EoUppRr2FvghjQYXX3yxplfqItX34osvkgD6W8grcf+dzn4cbBEk/J0c5wrHrOSg2qmkpyZRwpOJNy1btizhgfK+JLFK5tCJCcGU2vSyZcvOP/98vVqGUZFbCBVI7NK4cWM30Ve7fPny5Hfw9OajpHRpIjGZ+eyGnHHdEnZTiipQoEDccEiCMoEE4YRSrCExrtp6OcFhSjp66tSpeulOb86cOaUsj8PgORcsWJBIVEqLeJg1lxTN33nnHdJkGh53lxSYBi69uIWH1WZZ14tiQykxsIuXuhy6FBiwYCZYoI6rtqOAdA4I0zZt2owfP14SGkv39duJ+fPnOzukWDl+KQQMCxcuNAyOpZ8NkTY3nJcw4z8hfXi3lZPGdUs//vjjZ599phwrOcsoswRTDfSjY/WNuGrrhweBKd7O7863c+fOej2ivQMHDqxcuXL0MvFGyZIlPeVySSbldWXX2WefraTHErNmzVojpI8UI8ZOGkuR1n09TAkJpLBHSsfGTp0kShCYTpo0SW/EGTNmsKkiZuUduuCCC6pUqUL9pnjx4nHvgdgATj3b7NmzyX4j2TnYmY4ePXrTpk2aUWTma9WqpWFwuqTEWViu3ZklwfMfBr+FklBVrlw5ikySAlhS6nr33XelLkO6idoaUb5hStapdu3akkSy7iRQ27Zty9LH6gMbz55c1cyZM03S40OHDmWIdOBoWQpbtmzp/kIg61FaWlqpUqX69esnDYQ+ceJEzvqDYoGSwfzvCCmHh0uUSkTuWSRvyliNQ5VWfFyA8yjdU/htM7XfIW5+3zAlkeke726TA2J/vWbNGjfRvJ07d+5GjRpJ/EB/wIABhw4dkhiAeMOGDaXeokWLkjuTeh36l19+6TQ8ZylR5WFLnUvKfhSNlPpoYCp1SaBXyk8S0R9McWYNGjRQqkISnoX+4MGDyl4T4j333COxUWQikJB6o/RXXnmlefPm0UtPo02bNh6K51Laefja/3lknqpLaZmWXCbBAC5GqW3Gg6mEUW6vWbNmrO80Ah/4S+VYIq2RI0cqu2KJc+fOlb40ePvtt8fyuykEKsq8CVtG6Xu07uHp01ZqGDu1hC3JZQJfaV2WRMVOqqEYqi1J8OdNpY3I66+/TmpNmsOEXq1aNSkVz2puIiHKI/0+hC2dPhd79OhRwuioHHejU6dO7stT2I5EIiazS9jCCGXKlImVAExjiVAIHr766isaCR6RiFrtSERN90znD6ZSPclkg+KZ2HNJQctDcS4p0C9atMhpG563bt1KVl/JzPZOSY8SKc9G2+4G637Tpk3dFGWb0j+RT+IHb8XJkyeVUxhumamPSN+0VyJSSUQBKXigy9chqW3oZf3BVMqrJVhJI+Rt0aKF8rbnzJmjpOuJzjYuloeCYSzRTSEClgw6efLkuJmp48ePEzkkeOzatWvIkCEEi27Fou1IxMj9UDSiVhwd5W4o130pG0XE5R4buB2JGKktyfcHU6mcSB5KmsCEnpaWRmJcyWlYj/WM3bZtm4fiXObJk8dpSGe2gNKLQQL4zTffzJcvnzQ2FDoYWrlyZY4cOSQ3E4mYPm9p3WcKj6qlS5cmGPAQnUtJiNNrfo5ETNVWyvQHU6WIBIlA54EHHlAKId4laafs0hOlFVM/yul9WP4bvxS0cFHJy09169aNNPAl2j9WJcHXUd59lhxhkSJFLrroIjentOITNkjBunu4SdtcbaU0fzDdu3evUgp7YSXdhEjSXnKlzzzzjImEWB5pcZf0d0sgvmTNdVPcbSTjYKQv+bs5fbUJewjBn3jiCV+j9MwgXmKoWbOmuyvWvzq9uAmCB6d9as/+YCr96NFz2+a3RMq9ffv2Sn4gRaSo7IpLlPTZs2dP3LEwPProo5rtLd5uyZIl06ZNkxZKJJgf55xzTo8ePags1KtXzz0qEklolUQUpRCp8ufBpeeSsc7BC+k0TvnZH0zZRCs1ZiOspOuJV155ZfTPQsVyBvsGPnIIH6UNAZ4ShrgHMQP7+p+1P5ojywsIHnvsscAxACtv7969MSl3StujlbRKSnTPcOdSWvfdqzxTU6Jz+D3nEGHqS22PGlz6gyl7CMbEHvXr169SpUosXUPBJyGNfYmShzzUuHHjlF1xidKSfeLEif+nV+KO/5Nh3bp1cd+9vHnzUr8FZyyO/fr1q127Njf152jdiZpwhw4dFixYQEpy2LBhfuMlKQmvnFLCWcWKFaPfWqxRo4ZyLERN2ECvr0NSOxIxWjT8wZTFTlLu2WefNfxaJxKI8JYvX655QiQOpa/VMVxz1KlTRyq6Ll68GDepGevpgt/wf6KqXLny4MGD33jjjf379xOrrFixYuHChRiENw13O3bs2Hnz5pGlwvtSQSBwmjRpUuD/M8OX7TM9FQAABTxJREFUW5JgmilTpqhDlVZ84hDCBo9NAl9Kakt0z0T+YIqJpV8YFi9e3DCUJPtNnpUkiEeV6OXzzz8f7O8qUl+RKqUIp+LP2deBJoZIjYolccENNmzYkLela9euuNv777+/SZMmOK1ixYpJq0d0eNyG4XN15HzxxRdSBTsuTCWIO5L9nn2pHSvcH0wZL5Ui6eJhrF+//rrrrqOtPHLlyjVmzBicjeZHfISPuFLlcIjSd+oyZ87cvXv3DRs2SJLxZL7+9i9zOQdI5YkShDiX6Xb+5ZdflHNFIkarZHSsVJBznChLPwFAlNndCBemkYg/td2a0PYNU7yp9BdgEEfRnF6Siz179mQpzJ49O5lqLEIykqz1gQMHNN8DZDgHEaHGAezevZtAsG/fvtWrVyfGKPznf2zH2rpz587Ro0czXDrizisNhI7mZcuWZWtPOx0OXmOWpmB/TiJWPQltvHswV6tWjQCARuwhbb9iOU0o6e1N0Unj7ejlYDs1YsQI8HTkyJHDhw8TieODq1atSpf+YK0kHpB4nF+2gH42STgJKpNUm4ggWVv1e5eXX36ZwFESa0KnOkXiDMfDvZjwB+NhFkKFWrVqERdKEqS9iMQvoY1MLS8DHkQ5EE+xefNmZVcwol+1PbP49qaM//zzz+MiFTa/R6tWrTThI5UqXLVfmfCjLZJpaA+jzrVr15Lqqlu3ruZdMhIUw0RM0qdPH3JbmtjaGeTXLeEsKCY5Yz1nMMrhITqXeAGnEdbZr9qeeYPAFBHUh2bOnEkjrIM0pCZ2LFSoELmbAHPhn0BV4r+RcE9NuqNChQqsGNOnT2etcHf5bZOaBZdAnxTy8OHDyQP4lRCXnxiXt0vJBkadpT+2VwoVYjnThxIQpijXunVrgkIaCR48aWpR+iwBcSGbJL8TEchSjiJs9TvQhJ9id7t27ciNV6pUiaTp/Pnz2VabDIRn9erVxC233HILw1nlfQUSJFv+LXykfR7JMuUI1oTx48cru0jromfsQfik5J8yZUoss5sycuRI5UDDTHZwmKIECzGFPhqBj7lz55YsWZIcqkYCu58AX57fuHEjAGLF10gOpWvNmjVDhw6lalWiRIlIJFKuXLkWLVqQh4o+FTZ8Xbp0AY7kSslVkZMivO7fv/+yZcuoOPjVgbzsv4SPBFOye8oRTz31lJIOUQqOgSm9scfUeH9YhL1K7Cgo6QFTTEyCCZzxXtL2dZAQYLfevHlz3INmIHujAG8Cu37yYnrJmkkT6SIlN2fOnIEDB/IMnINwZcKECSzuPGOS/OxOEpF/eo5NyJs6JmNLSKzWpk0b3IND0Z+Je6gVgVGQqueklxwWxSp22dL7DU/0YO8PIMitpqWlHTt2LEq3jYxugRBg6piA2iDBFgn8li1bkjnyfE+Uy4kTJ9JVpEgR/BzVfGeU4ZnYlOwJiyb7NjbF0VHsOchJkaYlxqKkni1bNpZXae2LjjJvWM4UsUBoMHXuh5317NmzO3fuXKpUKQK16MFlx44d6aLI5HAGOLNosm9jUxwVS+2gaNGi5KgpSC5dujSATDskQ1ggZJhmiHu2SmY4C1iYZrhHdjoqbGF6Oj71DHfPFqYZ7pGdjgpbmPp/6nZEulvAwjTdTW4n9G8BC1P/NrMj0t0CFqbpbnI7oX8LWJj6t5kdke4WsDBNd5PbCf1bwMLUv83UIyw1iRawME2ica3osCxgYRqWJa2cJFrAwjSJxrWiw7KAhWlYlrRykmgBC9MkGteKDssCFqZhWVItx1JDsYCFaShmtEKSa4H/AQAA///gGR1OAAAABklEQVQDAGi0E0yUIq5nAAAAAElFTkSuQmCC" width="226" height="98" class="img_ev3q"></p>
<p>GMI Cloud 是一款面向生产级人工智能推理场景打造的原生 AI 基础设施平台。从无服务器应用程序接口到专属 GPU 集群，GMI Cloud 依托英伟达 GPU 平台，可提供稳定可控的性能、弹性扩容的算力以及高性价比的运行服务。</p>
<p>GMI Cloud 打造了全栈垂直整合的人工智能基础设施体系，覆盖推理接口、调度编排、算力计算至硬件设备的全链路环节。</p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="关于-karmada-用户组">关于 Karmada 用户组<a href="https://karmada.io/zh/blog/2026/04/24/gmi-cloud-officially-joined-the-karmada-adopter-group/gmi-cloud#%E5%85%B3%E4%BA%8E-karmada-%E7%94%A8%E6%88%B7%E7%BB%84" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p><strong>Karmada 用户组</strong>是一个由在其环境中成功采用 Karmada 的组织和用户组成的社区。成员分享他们的经验、最佳实践和反馈，以帮助改进 Karmada 并培育一个充满活力的协作生态系统。</p>
<p>成为 Karmada 用户组成员具有以下几个优势：</p>
<ul>
<li class=""><strong>社区认可</strong>：作为云原生多集群管理领域的领导者来展示您的组织，在 CNCF 和 Karmada 社区中获得知名度。</li>
<li class=""><strong>促进协作</strong>：与其他采用者建立联系，分享最佳实践，并在实际用例和解决方案上进行协作。</li>
<li class=""><strong>保持信息同步</strong>：及时接收重要更新通知，包括关键功能、错误修复和安全建议。</li>
<li class=""><strong>活动参与</strong>：受邀参与 Karmada 相关活动，包括 KubeCon + CloudNativeCon、网络研讨会和聚会。</li>
<li class=""><strong>职位发布</strong>：有机会在 Karmada 社区支持的职位公告板上发布与 Karmada 相关的职位空缺（目前暂不可用）。</li>
<li class=""><strong>扩展商业机会</strong>：与 Karmada 生态系统的其他成员建立潜在的商业联系和合作。</li>
</ul>
<p>在 <a href="https://github.com/karmada-io/community/tree/main/adopter-group" target="_blank" rel="noopener noreferrer" class="">GitHub 社区仓库</a> 中了解更多关于 Karmada 用户组的信息，并在 <a href="https://karmada.io/zh/adopters" target="_blank" rel="noopener noreferrer" class="">karmada.io/adopters</a> 查看完整的<code>公开</code>的采用者列表。</p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="加入-karmada-用户组">加入 Karmada 用户组<a href="https://karmada.io/zh/blog/2026/04/24/gmi-cloud-officially-joined-the-karmada-adopter-group/gmi-cloud#%E5%8A%A0%E5%85%A5-karmada-%E7%94%A8%E6%88%B7%E7%BB%84" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p><strong>Karmada 用户组</strong>对当前正在<strong>生产环境</strong>中使用 Karmada 的<strong>最终用户</strong>和<strong>供应商</strong>开放。这包括：</p>
<ul>
<li class=""><strong>最终用户</strong>：在其生产环境中运行 Karmada 的组织。</li>
<li class=""><strong>供应商</strong>：提供基于 Karmada 的产品或服务，并有客户在生产环境中使用这些产品或服务的公司。</li>
</ul>
<p>您是否在生产环境中使用 Karmada 并有兴趣加入 Karmada 用户组？</p>
<p>加入很简单——只需填写 <a href="https://github.com/karmada-io/karmada/issues/4540" target="_blank" rel="noopener noreferrer" class="">Karmada 用户组申请表单</a> 并提供您组织的信息。在您的申请经过审核和批准后，您的组织将被添加到 Karmada 用户组。</p>
<p>更多详情，请查看社区仓库中的 <a href="https://github.com/karmada-io/community/tree/main/adopter-group#how-to-join" target="_blank" rel="noopener noreferrer" class="">如何加入</a> 部分。</p>
<p>让我们一起发展 Karmada 社区和更广泛的多集群生态系统！</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Karmada v1.17 版本发布！新增工作负载亲和性支持]]></title>
            <link>https://karmada.io/zh/blog/2026/03/17/karmada-v1.17/karmada-v1.17</link>
            <guid>https://karmada.io/zh/blog/2026/03/17/karmada-v1.17/karmada-v1.17</guid>
            <pubDate>Tue, 17 Mar 2026 00:00:00 GMT</pubDate>
            <description><![CDATA[Karmada 是开放的多云多集群容器编排引擎，旨在帮助用户在多云环境下部署和运维业务应用。凭借兼容 Kubernetes 原生 API 的能力，Karmada 可以平滑迁移单集群工作负载，并且仍可保持与 Kubernetes 周边生态工具链协同。]]></description>
            <content:encoded><![CDATA[<p>Karmada 是开放的多云多集群容器编排引擎，旨在帮助用户在多云环境下部署和运维业务应用。凭借兼容 Kubernetes 原生 API 的能力，Karmada 可以平滑迁移单集群工作负载，并且仍可保持与 Kubernetes 周边生态工具链协同。</p>
<p><a href="https://github.com/karmada-io/karmada/blob/master/docs/CHANGELOG/CHANGELOG-1.17.md" target="_blank" rel="noopener noreferrer" class="">Karmada v1.17</a> 版本现已发布，本版本包含下列新增特性：</p>
<ul>
<li class="">工作负载亲和与反亲和调度</li>
<li class="">Dashboard v0.3.0 发布</li>
<li class="">持续的性能优化</li>
</ul>
<p>这些特性使 Karmada 在处理大规模、复杂的多集群场景时更加成熟和可靠。我们鼓励您升级到 v1.17.0，体验这些新功能带来的价值。</p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="新特性概览">新特性概览<a href="https://karmada.io/zh/blog/2026/03/17/karmada-v1.17/karmada-v1.17#%E6%96%B0%E7%89%B9%E6%80%A7%E6%A6%82%E8%A7%88" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<h3 class="anchor anchorTargetStickyNavbar_Vzrq" id="支持工作负载亲和与反亲和调度">支持工作负载亲和与反亲和调度<a href="https://karmada.io/zh/blog/2026/03/17/karmada-v1.17/karmada-v1.17#%E6%94%AF%E6%8C%81%E5%B7%A5%E4%BD%9C%E8%B4%9F%E8%BD%BD%E4%BA%B2%E5%92%8C%E4%B8%8E%E5%8F%8D%E4%BA%B2%E5%92%8C%E8%B0%83%E5%BA%A6" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h3>
<p>在多集群场景下，大量应用对工作负载之间的部署位置存在明确诉求，以实现高可用、低延迟、成本优化与运维隔离等目标。为满足这类精细化部署需求，本版本正式推出<strong>工作负载亲和与反亲和调度</strong>能力，让你能够精细控制工作负载在多集群间的拓扑关系。</p>
<h4 class="anchor anchorTargetStickyNavbar_Vzrq" id="工作负载亲和workload-affinity">工作负载亲和（Workload Affinity）<a href="https://karmada.io/zh/blog/2026/03/17/karmada-v1.17/karmada-v1.17#%E5%B7%A5%E4%BD%9C%E8%B4%9F%E8%BD%BD%E4%BA%B2%E5%92%8Cworkload-affinity" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h4>
<p>将存在关联的工作负载（如微服务与其缓存、分布式训练任务等）调度到同一集群。</p>
<ul>
<li class="">核心价值：降低跨集群网络延迟，显著提升性能敏感型应用的运行效率。</li>
<li class="">典型场景：服务与依赖组件同集群部署、训练任务组件就近调度。</li>
</ul>
<h4 class="anchor anchorTargetStickyNavbar_Vzrq" id="工作负载反亲和workload-anti-affinity">工作负载反亲和（Workload Anti-Affinity）<a href="https://karmada.io/zh/blog/2026/03/17/karmada-v1.17/karmada-v1.17#%E5%B7%A5%E4%BD%9C%E8%B4%9F%E8%BD%BD%E5%8F%8D%E4%BA%B2%E5%92%8Cworkload-anti-affinity" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h4>
<p>将同一逻辑组的工作负载分散到不同集群。</p>
<ul>
<li class="">核心价值：避免单集群故障导致关键应用整体不可用，强化多集群高可用保障。</li>
<li class="">典型场景：核心服务多集群分散、多副本跨集群容灾。</li>
</ul>
<h4 class="anchor anchorTargetStickyNavbar_Vzrq" id="使用方式">使用方式<a href="https://karmada.io/zh/blog/2026/03/17/karmada-v1.17/karmada-v1.17#%E4%BD%BF%E7%94%A8%E6%96%B9%E5%BC%8F" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h4>
<p>只需在 PropagationPolicy 中添加 workloadAffinity 配置，即可基于资源模板上的标签定义亲和组，实现工作负载的同集群部署或跨集群分散。<br>
<!-- -->让我们举一个<strong>工作负载亲和性</strong>的例子，假设您有一组训练任务，为了达到最佳的训练效果，您希望这一组训练任务的作业能在同一个集群上运行，您可以这样配置：</p>
<div class="language-yaml codeBlockContainer_Ckt0 theme-code-block" style="--prism-color:#393A34;--prism-background-color:#f6f8fa"><div class="codeBlockContent_QJqH"><pre tabindex="0" class="prism-code language-yaml codeBlock_bY9V thin-scrollbar" style="color:#393A34;background-color:#f6f8fa"><code class="codeBlockLines_e6Vv"><span class="token-line" style="color:#393A34"><span class="token key atrule" style="color:#00a4db">apiVersion</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> policy.karmada.io/v1alpha1</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">kind</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> PropagationPolicy</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">metadata</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">name</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> training</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">tasks</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">affinity</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">example</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">namespace</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> default</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">spec</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">resourceSelectors</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">apiVersion</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> batch/v1</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">kind</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> Job</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">labelSelector</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">        </span><span class="token key atrule" style="color:#00a4db">matchLabels</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">          </span><span class="token key atrule" style="color:#00a4db">workload.type</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> training</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">placement</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">spreadConstraints</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">maxGroups</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token number" style="color:#36acaa">1</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">        </span><span class="token key atrule" style="color:#00a4db">minGroups</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token number" style="color:#36acaa">1</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">clusterAffinity</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">clusterNames</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">        </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> member1</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">        </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> member2</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">        </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> member3</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">workloadAffinity</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">affinity</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">        </span><span class="token key atrule" style="color:#00a4db">groupByLabelKey</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> app.training</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">group</span><br></span></code></pre></div></div>
<p>**启用工作负载亲和与反亲和调度功能后，**Karmada 会将具有相同 <code>app.training-group</code> 标签值的训练任务调度到同一个集群。</p>
<p>再来看一个<strong>工作负载反亲和性</strong>的例子，假设您运行着对停机时间非常敏感的 Flink 数据处理任务，为了保证高可用，您会部署多套相同的 Flink 任务副本。为了避免因单个集群故障导致服务完全中断，您希望同一任务的多个副本分散到不同集群上运行，您可以这样配置：</p>
<div class="language-yaml codeBlockContainer_Ckt0 theme-code-block" style="--prism-color:#393A34;--prism-background-color:#f6f8fa"><div class="codeBlockContent_QJqH"><pre tabindex="0" class="prism-code language-yaml codeBlock_bY9V thin-scrollbar" style="color:#393A34;background-color:#f6f8fa"><code class="codeBlockLines_e6Vv"><span class="token-line" style="color:#393A34"><span class="token key atrule" style="color:#00a4db">apiVersion</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> policy.karmada.io/v1alpha1</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">kind</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> PropagationPolicy</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">metadata</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">name</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> flink</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">anti</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">affinity</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">example</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">namespace</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> default</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">spec</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">resourceSelectors</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">apiVersion</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> flink.apache.org/v1beta1</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">kind</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> FlinkDeployment</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">labelSelector</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">        </span><span class="token key atrule" style="color:#00a4db">matchLabels</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">          </span><span class="token key atrule" style="color:#00a4db">ha.enabled</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token string" style="color:#e3116c">"true"</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">placement</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">spreadConstraints</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">maxGroups</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token number" style="color:#36acaa">1</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">        </span><span class="token key atrule" style="color:#00a4db">minGroups</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token number" style="color:#36acaa">1</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">clusterAffinity</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">clusterNames</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">        </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> clusterA</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">        </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> clusterB</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">        </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> clusterC</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">workloadAffinity</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">antiAffinity</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">        </span><span class="token key atrule" style="color:#00a4db">groupByLabelKey</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> karmada.io/group</span><br></span></code></pre></div></div>
<p>**启用工作负载亲和与反亲和调度功能后，**Karmada 会将具有相同 <code>karmada.io/group</code> 标签值的 Flink 任务调度到不同集群，效果图如下：</p>
<p><img decoding="async" loading="lazy" alt="workload-anti-affinity" src="https://karmada.io/zh/assets/images/workload-anti-affinity-2b13dffcd050d48da92242bb6af37e9c.png" width="2030" height="1180" class="img_ev3q"></p>
<p>更多有关此功能的资料请参考：<a href="https://karmada.io/zh/docs/next/userguide/scheduling/propagation-policy/#workloadaffinity%E5%B7%A5%E4%BD%9C%E8%B4%9F%E8%BD%BD%E4%BA%B2%E5%92%8C%E6%80%A7" target="_blank" rel="noopener noreferrer" class="">工作负载亲和性</a>。</p>
<h3 class="anchor anchorTargetStickyNavbar_Vzrq" id="dashboard-v030-发布">Dashboard v0.3.0 发布<a href="https://karmada.io/zh/blog/2026/03/17/karmada-v1.17/karmada-v1.17#dashboard-v030-%E5%8F%91%E5%B8%83" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h3>
<p>Karmada Dashboard 是一款专为 Karmada 用户设计的图形化界面工具，旨在简化多集群管理的操作流程，提升用户体验。通过 Dashboard，用户可以直观地查看集群状态、资源分布以及任务执行情况，同时还能轻松完成配置调整和策略部署。</p>
<p>经过社区开发者的共同努力，Karmada Dashboard v0.3.0 正式发布！本次更新带来智能助手、成员集群管理等能力，多集群运维体验全面升级，更易用、更稳定、更智能！</p>
<p>Karmada Dashboard v0.3.0 主要功能包括：</p>
<ul>
<li class="">**智能化运维：**深度集成 MCP (Model Context Protocol) 与 LLM，推出智能聊天助手。支持自然语言交互，实时响应与工具能力扩展，让多集群管理更加智能！</li>
<li class="">**成员集群管理能力增强：**新增成员集群专用仪表盘，支持实时日志查看和终端交互，实现对子集群的精细化掌控。</li>
<li class="">**界面优化：**采用更现代化的 UI 组件 Ant Design v6 同时升级至 React 19，界面更加美观；优化数据请求与构建性能，提升交互体验与运行效率。</li>
<li class="">**安全性和稳定性增强：**引入全面的 E2E 测试框架维护基本功能特性，并自动升级安全依赖，减少安全风险。</li>
</ul>
<p>下面是 Karmada Dashboard v0.3.0 新版界面展示，可以在顶部导航栏上切换成员集群，无缝切换到成员集群：</p>
<p><img decoding="async" loading="lazy" alt="dashboard-UI" src="https://karmada.io/zh/assets/images/dashboard-UI-14bfa096db9f89a254952cbd2ecdd13b.png" width="3338" height="2050" class="img_ev3q"></p>
<p>更多有关 Karmada Dashboard 的资料请参考：<a href="https://github.com/karmada-io/dashboard" target="_blank" rel="noopener noreferrer" class="">Karmada Dashboard</a></p>
<h3 class="anchor anchorTargetStickyNavbar_Vzrq" id="持续的性能优化">持续的性能优化<a href="https://karmada.io/zh/blog/2026/03/17/karmada-v1.17/karmada-v1.17#%E6%8C%81%E7%BB%AD%E7%9A%84%E6%80%A7%E8%83%BD%E4%BC%98%E5%8C%96" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h3>
<p>在此版本中，性能优化团队继续增强 Karmada 的性能，对控制器进行了如下改进：</p>
<h4 class="anchor anchorTargetStickyNavbar_Vzrq" id="控制器优先级队列controllerpriorityqueue特性升级至-beta-并默认启用">控制器优先级队列（ControllerPriorityQueue）特性升级至 Beta 并默认启用<a href="https://karmada.io/zh/blog/2026/03/17/karmada-v1.17/karmada-v1.17#%E6%8E%A7%E5%88%B6%E5%99%A8%E4%BC%98%E5%85%88%E7%BA%A7%E9%98%9F%E5%88%97controllerpriorityqueue%E7%89%B9%E6%80%A7%E5%8D%87%E7%BA%A7%E8%87%B3-beta-%E5%B9%B6%E9%BB%98%E8%AE%A4%E5%90%AF%E7%94%A8" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h4>
<p><code>控制器优先级队列</code>特性自 v1.15 版本首次推出后，历经两个版本的持续迭代与严格测试，能力已趋于成熟，本次版本正式将其升级为 Beta 版，并默认开启该特性。<br>
<!-- -->基于此特性，Karmada 控制器能够在重启或切主后<strong>立即响应并优先处理</strong>用户触发的资源变更，从而显著缩短服务重启和计划性升级过程中的停机时间。</p>
<h4 class="anchor anchorTargetStickyNavbar_Vzrq" id="依赖资源分发能力性能优化">依赖资源分发能力性能优化<a href="https://karmada.io/zh/blog/2026/03/17/karmada-v1.17/karmada-v1.17#%E4%BE%9D%E8%B5%96%E8%B5%84%E6%BA%90%E5%88%86%E5%8F%91%E8%83%BD%E5%8A%9B%E6%80%A7%E8%83%BD%E4%BC%98%E5%8C%96" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h4>
<p>通过消减并发更新依赖资源相关的 ResourceBinding 所产生的 API 冲突，依赖资源跟随分发的效率得到了显著的提升。<br>
<!-- -->测试环境包括 30,000 个 Workload 及其 PropagationPolicy，以及 30,000 个 ConfigMap 作为 Workload 依赖的资源。该优化将控制器首次启动的队列处理时间从20 分钟以上大幅缩短至约 5 分钟，显著提升了大规模集群下的系统响应速度。</p>
<p>有关详细的优化内容和测试报告，请参阅 PR: <a href="https://github.com/karmada-io/karmada/pull/7153" target="_blank" rel="noopener noreferrer" class="">[Performance] optimize the mechanism of create or update dependencies-distribute resourcebinding</a></p>
<h1>致谢贡献者</h1>
<p>Karmada v1.17 版本包含了来自 32 位贡献者的 264 次代码提交，在此对各位贡献者表示由衷的感谢：</p>
<table><thead><tr><th>^-^</th><th>^-^</th><th>^-^</th></tr></thead><tbody><tr><td>@7h3-3mp7y-m4n</td><td>@Abhay349</td><td>@abhinav-1305</td></tr><tr><td>@AbhinavPInamdar</td><td>@Ady0333</td><td>@Aman-Cool</td></tr><tr><td>@Arhell</td><td>@arnavgogia20</td><td>@CharlesQQ</td></tr><tr><td>@cmontemuino</td><td>@dahuo98</td><td>@FAUST-BENCHOU</td></tr><tr><td>@gmarav05</td><td>@goyalpalak18</td><td>@jabellard</td></tr><tr><td>@kajal-jotwani</td><td>@LivingCcj</td><td>@mohamedawnallah</td></tr><tr><td>@mszacillo</td><td>@RainbowMango</td><td>@rayo1uo</td></tr><tr><td>@seanlaii</td><td>@SunsetB612</td><td>@suresh-subramanian2013</td></tr><tr><td>@vie-serendipity</td><td>@warjiang</td><td>@XiShanYongYe-Chang</td></tr><tr><td>@yaten2302</td><td>@yoursanonymous</td><td>@zach593</td></tr><tr><td>@zhengjr9</td><td>@zhzhuang-zju</td><td></td></tr></tbody></table>
<p><img decoding="async" loading="lazy" alt="karmada v1.17 contributors" src="https://karmada.io/zh/assets/images/contributors-0814a117a69e3a80cc3c7e926f79eb1a.png" width="585" height="607" class="img_ev3q"></p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[欢迎松科智能加入 Karmada 用户组]]></title>
            <link>https://karmada.io/zh/blog/2026/03/17/songke-officially-joined-the-karmada-adopter-group/songke</link>
            <guid>https://karmada.io/zh/blog/2026/03/17/songke-officially-joined-the-karmada-adopter-group/songke</guid>
            <pubDate>Tue, 17 Mar 2026 00:00:00 GMT</pubDate>
            <description><![CDATA[Karmada 非常高兴地宣布 松科智能 正式加入 Karmada 用户组，成为社区的重要成员。作为云原生计算基金会（CNCF）旗下的项目，Karmada 致力于为用户提供强大的多集群管理和调度能力，帮助企业在复杂的分布式环境中实现高效的应用部署和管理。松科智能的加入将进一步加强 Karmada 社区，为项目的持续创新带来新的活力，标志着我们社区发展和 Karmada 在多样化生产环境中采用的又一个重要里程碑。]]></description>
            <content:encoded><![CDATA[<p>Karmada 非常高兴地宣布 <strong><a href="https://www.gdsoke.com/" target="_blank" rel="noopener noreferrer" class="">松科智能</a></strong> 正式加入 Karmada 用户组，成为社区的重要成员。作为云原生计算基金会（CNCF）旗下的项目，Karmada 致力于为用户提供强大的多集群管理和调度能力，帮助企业在复杂的分布式环境中实现高效的应用部署和管理。<strong>松科智能</strong>的加入将进一步加强 Karmada 社区，为项目的持续创新带来新的活力，标志着我们社区发展和 Karmada 在多样化生产环境中采用的又一个重要里程碑。</p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="关于松科智能">关于松科智能<a href="https://karmada.io/zh/blog/2026/03/17/songke-officially-joined-the-karmada-adopter-group/songke#%E5%85%B3%E4%BA%8E%E6%9D%BE%E7%A7%91%E6%99%BA%E8%83%BD" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p><img decoding="async" loading="lazy" alt="member logo" src="data:image/png;base64,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" width="144" height="144" class="img_ev3q"></p>
<p>广东松科智能科技有限公司成立于 2012 年，位于松山湖高新区大学创新园区，专注于深度学习神经网络的软硬件产品研发，致力研发一站式新一代人工智能教学平台，为培训机构、中小学、职高本科等提供人工智能人才培养的整体解决方案，包括有<!-- -->:AI<!-- -->案例&amp;课程体系、AI训练与编程平台、AI教具、AI教育机器人等软硬件和课程内容。</p>
<p>松科TPU&amp;AMC(AI 小脑)作为人工智能学习、开发、部署及竞赛的核心部件，为智能体的构建提供算力基础。其模块化、成本低、可迁移等特性，为普及AI项目实践教学提供保障，也为未来AI人才落地培养创造空间。</p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="关于-karmada-用户组">关于 Karmada 用户组<a href="https://karmada.io/zh/blog/2026/03/17/songke-officially-joined-the-karmada-adopter-group/songke#%E5%85%B3%E4%BA%8E-karmada-%E7%94%A8%E6%88%B7%E7%BB%84" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p><strong>Karmada 用户组</strong>是一个由在其环境中成功采用 Karmada 的组织和用户组成的社区。成员分享他们的经验、最佳实践和反馈，以帮助改进 Karmada 并培育一个充满活力的协作生态系统。</p>
<p>成为 Karmada 用户组成员具有以下几个优势：</p>
<ul>
<li class=""><strong>社区认可</strong>：作为云原生多集群管理领域的领导者来展示您的组织，在 CNCF 和 Karmada 社区中获得知名度。</li>
<li class=""><strong>促进协作</strong>：与其他采用者建立联系，分享最佳实践，并在实际用例和解决方案上进行协作。</li>
<li class=""><strong>保持信息同步</strong>：及时接收重要更新通知，包括关键功能、错误修复和安全建议。</li>
<li class=""><strong>活动参与</strong>：受邀参与 Karmada 相关活动，包括 KubeCon + CloudNativeCon、网络研讨会和聚会。</li>
<li class=""><strong>职位发布</strong>：有机会在 Karmada 社区支持的职位公告板上发布与 Karmada 相关的职位空缺（目前暂不可用）。</li>
<li class=""><strong>扩展商业机会</strong>：与 Karmada 生态系统的其他成员建立潜在的商业联系和合作。</li>
</ul>
<p>在 <a href="https://github.com/karmada-io/community/tree/main/adopter-group" target="_blank" rel="noopener noreferrer" class="">GitHub 社区仓库</a> 中了解更多关于 Karmada 用户组的信息，并在 <a href="https://karmada.io/zh/adopters" target="_blank" rel="noopener noreferrer" class="">karmada.io/adopters</a> 查看完整的<code>公开</code>的采用者列表。</p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="加入-karmada-用户组">加入 Karmada 用户组<a href="https://karmada.io/zh/blog/2026/03/17/songke-officially-joined-the-karmada-adopter-group/songke#%E5%8A%A0%E5%85%A5-karmada-%E7%94%A8%E6%88%B7%E7%BB%84" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p><strong>Karmada 用户组</strong>对当前正在<strong>生产环境</strong>中使用 Karmada 的<strong>最终用户</strong>和<strong>供应商</strong>开放。这包括：</p>
<ul>
<li class=""><strong>最终用户</strong>：在其生产环境中运行 Karmada 的组织。</li>
<li class=""><strong>供应商</strong>：提供基于 Karmada 的产品或服务，并有客户在生产环境中使用这些产品或服务的公司。</li>
</ul>
<p>您是否在生产环境中使用 Karmada 并有兴趣加入 Karmada 用户组？</p>
<p>加入很简单——只需填写 <a href="https://github.com/karmada-io/karmada/issues/4540" target="_blank" rel="noopener noreferrer" class="">Karmada 用户组申请表单</a> 并提供您组织的信息。在您的申请经过审核和批准后，您的组织将被添加到 Karmada 用户组。</p>
<p>更多详情，请查看社区仓库中的 <a href="https://github.com/karmada-io/community/tree/main/adopter-group#how-to-join" target="_blank" rel="noopener noreferrer" class="">如何加入</a> 部分。</p>
<p>让我们一起发展 Karmada 社区和更广泛的多集群生态系统！</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[欢迎商汤加入 Karmada 用户组]]></title>
            <link>https://karmada.io/zh/blog/2025/12/30/sensetime-officially-joined-the-karmada-adopter-group/sensetime</link>
            <guid>https://karmada.io/zh/blog/2025/12/30/sensetime-officially-joined-the-karmada-adopter-group/sensetime</guid>
            <pubDate>Tue, 30 Dec 2025 00:00:00 GMT</pubDate>
            <description><![CDATA[Karmada 非常高兴地宣布 商汤 正式加入 Karmada 用户组，成为社区的重要成员。作为云原生计算基金会（CNCF）旗下的项目，Karmada 致力于为用户提供强大的多集群管理和调度能力，帮助企业在复杂的分布式环境中实现高效的应用部署和管理。商汤的加入将进一步加强 Karmada 社区，为项目的持续创新带来新的活力，标志着我们社区发展和 Karmada 在多样化生产环境中采用的又一个重要里程碑。]]></description>
            <content:encoded><![CDATA[<p>Karmada 非常高兴地宣布 <strong><a href="https://www.sensetime.com/" target="_blank" rel="noopener noreferrer" class="">商汤</a></strong> 正式加入 Karmada 用户组，成为社区的重要成员。作为云原生计算基金会（CNCF）旗下的项目，Karmada 致力于为用户提供强大的多集群管理和调度能力，帮助企业在复杂的分布式环境中实现高效的应用部署和管理。<strong>商汤</strong>的加入将进一步加强 Karmada 社区，为项目的持续创新带来新的活力，标志着我们社区发展和 Karmada 在多样化生产环境中采用的又一个重要里程碑。</p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="关于商汤">关于商汤<a href="https://karmada.io/zh/blog/2025/12/30/sensetime-officially-joined-the-karmada-adopter-group/sensetime#%E5%85%B3%E4%BA%8E%E5%95%86%E6%B1%A4" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p><img decoding="async" loading="lazy" alt="member logo" src="https://karmada.io/zh/assets/images/sensetime-1909b96adda8ab1c46c9991914a95f5d.png" width="310" height="88" class="img_ev3q"></p>
<p>商汤科技坚持"AI基础设施（大装置）-大模型（日日新）-应用"三位一体战略。商汤大装置作为商汤科技前瞻打造的高效率、低成本、规模化的新一代AI基础设施，以大模型开发、生成、应用为核心，赋能人工智能生产新范式。通过与大模型迭代的联合调优，打造"最懂大模型的AI基础设施"。</p>
<p>2022年，作为商汤大装置重要载体的人工智能计算中心（上海临港AIDC）正式投入运营，成为亚洲最大人工智能计算中心之一。过去三年，商汤持续投入建设并自持全国首个5A级智算中心上海临港AIDC，以及通过运营模式将算力总规模提升至32,000 PFlops（截至2025年12月）。强大算力可支撑超过20多个千亿参数超大模型同时训练，并支持万亿参数大模型的全生命周期生成。</p>
<p>作为行业先行者，商汤大装置已构建起"算力-平台-方案-服务"的端到端系统化能力，既为商汤日日新大模型的迭代提供了核心动力，也快速承接了各行业客户的大模型创新需求，持续赋能AIGC、具身智能、AI4S、产业智能化等相关领域创新。</p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="关于-karmada-用户组">关于 Karmada 用户组<a href="https://karmada.io/zh/blog/2025/12/30/sensetime-officially-joined-the-karmada-adopter-group/sensetime#%E5%85%B3%E4%BA%8E-karmada-%E7%94%A8%E6%88%B7%E7%BB%84" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p><strong>Karmada 用户组</strong>是一个由在其环境中成功采用 Karmada 的组织和用户组成的社区。成员分享他们的经验、最佳实践和反馈，以帮助改进 Karmada 并培育一个充满活力的协作生态系统。</p>
<p>成为 Karmada 用户组成员具有以下几个优势：</p>
<ul>
<li class=""><strong>社区认可</strong>：作为云原生多集群管理领域的领导者来展示您的组织，在 CNCF 和 Karmada 社区中获得知名度。</li>
<li class=""><strong>促进协作</strong>：与其他采用者建立联系，分享最佳实践，并在实际用例和解决方案上进行协作。</li>
<li class=""><strong>保持信息同步</strong>：及时接收重要更新通知，包括关键功能、错误修复和安全建议。</li>
<li class=""><strong>活动参与</strong>：受邀参与 Karmada 相关活动，包括 KubeCon + CloudNativeCon、网络研讨会和聚会。</li>
<li class=""><strong>职位发布</strong>：有机会在 Karmada 社区支持的职位公告板上发布与 Karmada 相关的职位空缺（目前暂不可用）。</li>
<li class=""><strong>扩展商业机会</strong>：与 Karmada 生态系统的其他成员建立潜在的商业联系和合作。</li>
</ul>
<p>在 <a href="https://github.com/karmada-io/community/tree/main/adopter-group" target="_blank" rel="noopener noreferrer" class="">GitHub 社区仓库</a> 中了解更多关于 Karmada 用户组的信息，并在 <a href="https://karmada.io/zh/adopters" target="_blank" rel="noopener noreferrer" class="">karmada.io/adopters</a> 查看完整的<code>公开</code>的采用者列表。</p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="加入-karmada-用户组">加入 Karmada 用户组<a href="https://karmada.io/zh/blog/2025/12/30/sensetime-officially-joined-the-karmada-adopter-group/sensetime#%E5%8A%A0%E5%85%A5-karmada-%E7%94%A8%E6%88%B7%E7%BB%84" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p><strong>Karmada 用户组</strong>对当前正在<strong>生产环境</strong>中使用 Karmada 的<strong>最终用户</strong>和<strong>供应商</strong>开放。这包括：</p>
<ul>
<li class=""><strong>最终用户</strong>：在其生产环境中运行 Karmada 的组织。</li>
<li class=""><strong>供应商</strong>：提供基于 Karmada 的产品或服务，并有客户在生产环境中使用这些产品或服务的公司。</li>
</ul>
<p>您是否在生产环境中使用 Karmada 并有兴趣加入 Karmada 用户组？</p>
<p>加入很简单——只需填写 <a href="https://github.com/karmada-io/karmada/issues/4540" target="_blank" rel="noopener noreferrer" class="">Karmada 用户组申请表单</a> 并提供您组织的信息。在您的申请经过审核和批准后，您的组织将被添加到 Karmada 用户组。</p>
<p>更多详情，请查看社区仓库中的 <a href="https://github.com/karmada-io/community/tree/main/adopter-group#how-to-join" target="_blank" rel="noopener noreferrer" class="">如何加入</a> 部分。</p>
<p>让我们一起发展 Karmada 社区和更广泛的多集群生态系统！</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Karmada v1.16 版本发布！支持多模板工作负载调度]]></title>
            <link>https://karmada.io/zh/blog/2025/12/05/karmada-v1.16/karmada-v1.16</link>
            <guid>https://karmada.io/zh/blog/2025/12/05/karmada-v1.16/karmada-v1.16</guid>
            <pubDate>Fri, 05 Dec 2025 00:00:00 GMT</pubDate>
            <description><![CDATA[Karmada 是开放的多云多集群容器编排引擎，旨在帮助用户在多云环境下部署和运维业务应用。凭借兼容 Kubernetes 原生 API 的能力，Karmada 可以平滑迁移单集群工作负载，并且仍可保持与 Kubernetes 周边生态工具链协同。]]></description>
            <content:encoded><![CDATA[<p>Karmada 是开放的多云多集群容器编排引擎，旨在帮助用户在多云环境下部署和运维业务应用。凭借兼容 Kubernetes 原生 API 的能力，Karmada 可以平滑迁移单集群工作负载，并且仍可保持与 Kubernetes 周边生态工具链协同。</p>
<p><a href="https://github.com/karmada-io/karmada/blob/master/docs/CHANGELOG/CHANGELOG-1.16.md" target="_blank" rel="noopener noreferrer" class="">Karmada v1.16</a> 版本现已发布，本版本包含下列新增特性：</p>
<ul>
<li class="">支持多模板工作负载调度</li>
<li class="">使用 Webster 算法增强副本分配</li>
<li class="">驱逐队列速率限制</li>
<li class="">持续的性能优化</li>
</ul>
<p>这些特性使 Karmada 在处理大规模、复杂的多集群场景时更加成熟和可靠。我们鼓励您升级到 v1.16.0，体验这些新功能带来的价值。</p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="新特性概览">新特性概览<a href="https://karmada.io/zh/blog/2025/12/05/karmada-v1.16/karmada-v1.16#%E6%96%B0%E7%89%B9%E6%80%A7%E6%A6%82%E8%A7%88" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<h3 class="anchor anchorTargetStickyNavbar_Vzrq" id="支持多模板工作负载调度">支持多模板工作负载调度<a href="https://karmada.io/zh/blog/2025/12/05/karmada-v1.16/karmada-v1.16#%E6%94%AF%E6%8C%81%E5%A4%9A%E6%A8%A1%E6%9D%BF%E5%B7%A5%E4%BD%9C%E8%B4%9F%E8%BD%BD%E8%B0%83%E5%BA%A6" class="hash-link" aria-label="跳转到头部" title="跳转到��头部" translate="no">​</a></h3>
<p>当前许多AI/大数据应用由多个组件构成，这些组件之间相互协作以完成复杂的计算任务。例如：</p>
<ul>
<li class="">FlinkDeployment 包含 JobManager 和 TaskManager，</li>
<li class="">SparkApplication 包含 Driver 和 Executor，</li>
<li class="">RayCluster 包含 Head 和 Worker 节点。</li>
</ul>
<p>在 Karmada v1.16.0 中，我们引入了<strong>多模板调度</strong>，这是一项全新的能力，使得 Karmada 能够将由多个相互关联组件组成的多模板工作负载完整且统一地调度到具有充足资源的单个成员集群中。</p>
<p>此功能建立在 v1.15 版本引入的<a href="https://karmada.io/blog/2025/09/05/karmada-v1.15/karmada-v1.15#precise-resource-awareness-for-multi-template-workloads" target="_blank" rel="noopener noreferrer" class="">多模板工作负载资源精确感知</a>功能之上，该支持使 Karmada 能够
准确理解复杂工作负载的资源拓扑。在 v1.16 中，调度器利用这些信息来：</p>
<ul>
<li class="">基于 ResourceQuota 限制来估算成员集群可以容纳的完整的多模板工作负载数；</li>
<li class="">利用成员集群中实际节点的可用资源来预测工作负载的可调度性。</li>
</ul>
<p>当前版本为以下多模板工作负载类型提供了内置的资源解释器：</p>
<ul>
<li class="">FlinkDeployment (flink.apache.org/v1beta1)</li>
<li class="">SparkApplication (sparkoperator.k8s.io/v1beta2)</li>
<li class="">Job (batch.volcano.sh/v1alpha1)</li>
<li class="">MPIJob (kubeflow.org/v2beta1)</li>
<li class="">RayCluster (ray.io/v1)</li>
<li class="">RayJob (ray.io/v1)</li>
<li class="">TFJob (kubeflow.org/v1)</li>
<li class="">PyTorchJob (kubeflow.org/v1)</li>
</ul>
<p>如果您使用的是其他的自定义多模板工作负载，也可以通过扩展 Karmada 的资源解释器来支持它们。</p>
<p>让我们举个简单的例子，假设您有一个 FlinkDeployment，其资源配置如下：</p>
<div class="language-yaml codeBlockContainer_Ckt0 theme-code-block" style="--prism-color:#393A34;--prism-background-color:#f6f8fa"><div class="codeBlockContent_QJqH"><pre tabindex="0" class="prism-code language-yaml codeBlock_bY9V thin-scrollbar" style="color:#393A34;background-color:#f6f8fa"><code class="codeBlockLines_e6Vv"><span class="token-line" style="color:#393A34"><span class="token key atrule" style="color:#00a4db">apiVersion</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> flink.apache.org/v1beta1</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">kind</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> FlinkDeployment</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">metadata</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">name</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> flink</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">example</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">spec</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">jobManager</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">replicas</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token number" style="color:#36acaa">1</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">resource</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">cpu</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token number" style="color:#36acaa">1</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">memory</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token string" style="color:#e3116c">"1024m"</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">taskManager</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">replicas</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token number" style="color:#36acaa">2</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">resource</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">cpu</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token number" style="color:#36acaa">2</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">memory</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token string" style="color:#e3116c">"2048m"</span><br></span></code></pre></div></div>
<p>启用多模板调度功能后，Karmada 会：</p>
<ol>
<li class="">通过资源解释器准确解析出 JobManager 和 TaskManager 的资源需求；</li>
<li class="">评估每个成员集群是否有足够资源容纳完整的 FlinkDeployment（1 个 JobManager + 2 个 TaskManager）；</li>
<li class="">将整个 FlinkDeployment 调度到单个满足条件的集群。</li>
</ol>
<p>此功能的发布标志着 Karmada 在支持AI/大数据应用方面迈出了重要一步——将精准的资源解释、配额感知计算和跨集群调度融合在一个统一的框架中。</p>
<h3 class="anchor anchorTargetStickyNavbar_Vzrq" id="使用-webster-算法增强副本分配">使用 Webster 算法增强副本分配<a href="https://karmada.io/zh/blog/2025/12/05/karmada-v1.16/karmada-v1.16#%E4%BD%BF%E7%94%A8-webster-%E7%AE%97%E6%B3%95%E5%A2%9E%E5%BC%BA%E5%89%AF%E6%9C%AC%E5%88%86%E9%85%8D" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h3>
<p>Karmada 支持多种副本调度策略，如 DynamicWeight、Aggregated 和 StaticWeight，用于在成员集群之间分配工作负载的副本。这些策略的核心在于将集群权重转化为实际副本数量的算法。</p>
<p>在之前的版本中，副本分配算法存在一定的局限性：</p>
<ul>
<li class="">非单调性：当总副本数增加时，某些集群可能意外地获得更少的副本.</li>
<li class="">缺乏强幂等性：相同的输入可能产生不同的输出.</li>
<li class="">不公平的余数分配：在具有相同权重的集群之间分配剩余副本时缺乏合理的优先级策略。</li>
</ul>
<p>在当前版本中，我们引入了 Webster 方法（也称为 Sainte-Laguë 方法）来改进跨集群调度期间的副本分配。通过采用 Webster 算法，Karmada 现在实现了：</p>
<ul>
<li class="">单调副本分配：增加总副本数绝不会导致任何集群丢失副本，确保行为一致且直观。</li>
<li class="">剩余副本的公平处理：在权重相等的集群间分配副本时，优先考虑当前副本数较少的集群。这种“小优先”方式有助于促进均衡部署，更好地满足高可用性（HA）需求。</li>
</ul>
<p>此次更新增强了跨集群工作负载分配的稳定性、公平性和可预测性，使多集群环境中的副本调度更加稳健。</p>
<h3 class="anchor anchorTargetStickyNavbar_Vzrq" id="驱逐队列速率限制">驱逐队列速率限制<a href="https://karmada.io/zh/blog/2025/12/05/karmada-v1.16/karmada-v1.16#%E9%A9%B1%E9%80%90%E9%98%9F%E5%88%97%E9%80%9F%E7%8E%87%E9%99%90%E5%88%B6" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h3>
<p>在多集群环境中，当集群发生故障时，资源需要从故障集群中驱逐并重新调度到健康的集群。如果多个集群同时或在短时间内相继发生故障，大量的驱逐和重新调度操作可能会使健康集群和控制平面不堪重负，进而导致级联故障。</p>
<p>此版本引入了具有速率限制功能的驱逐队列，用于 Karmada 污点管理器。驱逐队列通过可配置的固定速率参数来控制资源驱逐速率，从而增强故障迁移机制。该实现还提供了用于监控驱逐过程的指标，提高了整体系统的可观测性。</p>
<p>这个特性在以下场景特别有用：</p>
<ul>
<li class="">您需要在大规模故障期间防止级联故障，确保系统不会因为过多的驱逐操作而不堪重负。</li>
<li class="">您希望根据不同环境的特性配置驱逐行为。例如，在生产环境中使用较低的驱逐速率，在开发或测试环境中使用较高的速率。</li>
<li class="">您需要监控驱逐队列的性能，包括待处理驱逐的数量、处理延迟以及成功/失败率，以便调整配置和响应运维问题。</li>
</ul>
<p>驱逐队列的核心特性包括：</p>
<ul>
<li class="">可配置的固定速率限制：通过 --eviction-rate 命令行参数配置每秒驱逐速率。示例：设置每 2 秒最多驱逐 1 个资源：--eviction-rate=0.5。</li>
<li class="">完善的指标支持：提供队列深度、资源类型、处理延迟、成功/失败率等指标，便于监控和故障排查。
通过引入速率限制机制，管理员可以更好地控制集群故障迁移期间的资源调度速率，在保障服务稳定性的同时，提升资源调度的灵活性和效率。</li>
</ul>
<p>有关此功能的详细使用方法，请参阅<a href="https://karmada.io/docs/next/userguide/failover/cluster-failover/#configuring-eviction-rate-limiting" target="_blank" rel="noopener noreferrer" class="">用户指南：配置驱逐速率限制</a>。</p>
<h3 class="anchor anchorTargetStickyNavbar_Vzrq" id="持续的性能优化">持续的性能优化<a href="https://karmada.io/zh/blog/2025/12/05/karmada-v1.16/karmada-v1.16#%E6%8C%81%E7%BB%AD%E7%9A%84%E6%80%A7%E8%83%BD%E4%BC%98%E5%8C%96" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h3>
<p>在此版本中，性能优化团队继续增强 Karmada 的性能，对控制器进行了重大改进。</p>
<p>在 v1.15 中，我们引入了 <a href="https://github.com/kubernetes-sigs/controller-runtime/issues/2374" target="_blank" rel="noopener noreferrer" class="">controller-runtime 优先级队列</a>，它允许基于 controller-runtime 构建的控制器在重启或主从切换后优先处理最新的变更，从而显著减少服务重启和故障转移期间的停机时间。</p>
<p>在 v1.16 中，我们扩展了这一能力。对于不是基于 controller-runtime 构建的控制器（如 detector controller），我们通过为所有使用异步 worker 的控制器启用优先级队列功能，使它们也能享受到这一优化。</p>
<p>测试环境包括 5,000 个 Deployments 及其 PropagationPolicy，并在 karmada-controller-manager 组件中启用了 ControllerPriorityQueue 特性开关。在 karmada-controller-manager 组件重启后、工作队列中仍有大量待处理事件的情况下，手动修改一个 Deployment，其更新事件仍能被控制器快速处理并被同步到成员集群。</p>
<p>这些测试结果证明，Karmada 控制器的性能在 v1.16 中得到了很大的提升。未来，我们将继续对控制器和调度器进行系统性的性能优化。</p>
<p>有关详细的进展和测试报告，请参阅 PR: <a href="https://github.com/karmada-io/karmada/pull/6965" target="_blank" rel="noopener noreferrer" class="">[Performance] enable asyncPriorityWorker in all controllers</a>。</p>
<h1>致谢贡献者</h1>
<p>Karmada v1.16 版本包含了来自 30 位贡献者的 263 次代码提交，在此对各位贡献者表示由衷的感谢：</p>
<table><thead><tr><th>^-^</th><th>^-^</th><th>^-^</th></tr></thead><tbody><tr><td>@7h3-3mp7y-m4n</td><td>@baiyutang</td><td>@dekaihu</td></tr><tr><td>@devarsh10</td><td>@ikaven1024</td><td>@jabellard</td></tr><tr><td>@karmada-bot</td><td>@kasanatte</td><td>@Kexin2000</td></tr><tr><td>@liaolecheng</td><td>@LivingCcj</td><td>@luyb177</td></tr><tr><td>@mszacillo</td><td>@owenowenisme</td><td>@pokerfaceSad</td></tr><tr><td>@RainbowMango</td><td>@rayo1uo</td><td>@rohan-019</td></tr><tr><td>@ryanwuer</td><td>@satvik2131</td><td>@seanlaii</td></tr><tr><td>@ssenecal-modular</td><td>@SunsetB612</td><td>@vie-serendipity</td></tr><tr><td>@vsha96</td><td>@warjiang</td><td>@whosefriendA</td></tr><tr><td>@XiShanYongYe-Chang</td><td>@zach593</td><td>@zhzhuang-zju</td></tr></tbody></table>
<p><img decoding="async" loading="lazy" alt="karmada v1.16 contributors" src="https://karmada.io/zh/assets/images/contributors-69167482c558e3c9dcc937599da0e81e.png" width="593" height="507" class="img_ev3q"></p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[欢迎共绩科技加入 Karmada 用户组]]></title>
            <link>https://karmada.io/zh/blog/2025/11/18/gongjikeji-officially-joined-the-karmada-adopter-group/gongjikeji</link>
            <guid>https://karmada.io/zh/blog/2025/11/18/gongjikeji-officially-joined-the-karmada-adopter-group/gongjikeji</guid>
            <pubDate>Tue, 18 Nov 2025 00:00:00 GMT</pubDate>
            <description><![CDATA[Karmada 非常高兴地宣布 共绩科技 正式加入 Karmada 用户组，成为社区的重要成员。作为云原生计算基金会（CNCF）旗下的项目，Karmada 致力于为用户提供强大的多集群管理和调度能力，帮助企业在复杂的分布式环境中实现高效的应用部署和管理。共绩科技的加入将进一步加强 Karmada 社区，为项目的持续创新带来新的活力，标志着我们社区发展和 Karmada 在多样化生产环境中采用的又一个重要里程碑。]]></description>
            <content:encoded><![CDATA[<p>Karmada 非常高兴地宣布 <strong><a href="https://www.suanli.cn/" target="_blank" rel="noopener noreferrer" class="">共绩科技</a></strong> 正式加入 Karmada 用户组，成为社区的重要成员。作为云原生计算基金会（CNCF）旗下的项目，Karmada 致力于为用户提供强大的多集群管理和调度能力，帮助企业在复杂的分布式环境中实现高效的应用部署和管理。<strong>共绩科技</strong>的加入将进一步加强 Karmada 社区，为项目的持续创新带来新的活力，标志着我们社区发展和 Karmada 在多样化生产环境中采用的又一个重要里程碑。</p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="关于共绩科技">关于共绩科技<a href="https://karmada.io/zh/blog/2025/11/18/gongjikeji-officially-joined-the-karmada-adopter-group/gongjikeji#%E5%85%B3%E4%BA%8E%E5%85%B1%E7%BB%A9%E7%A7%91%E6%8A%80" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p><img decoding="async" loading="lazy" alt="member logo" src="https://karmada.io/zh/assets/images/gongjikeji-2849e75924e3a593bafe27b20653aaa7.png" width="512" height="96" class="img_ev3q"></p>
<p>共绩科技 2023 年成立于清华，专注于构建融合算力与电力的智能调度网络，旨在提供平价、可靠、绿色的算力服务，使 AI 技术真正普及至每个人。通过精细调度算法，已为多家 AIGC 领军企业及科研机构提供高效算力，目标提升资源利用率 60%。作为一群清华背景的年轻创新者，他们在 2023 年秋季获得奇绩创坛等投资，正引领算力革命，开启普惠科技新篇章。</p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="关于-karmada-用户组">关于 Karmada 用户组<a href="https://karmada.io/zh/blog/2025/11/18/gongjikeji-officially-joined-the-karmada-adopter-group/gongjikeji#%E5%85%B3%E4%BA%8E-karmada-%E7%94%A8%E6%88%B7%E7%BB%84" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p><strong>Karmada 用户组</strong>是一个由在其环境中成功采用 Karmada 的组织和用户组成的社区。成员分享他们的经验、最佳实践和反馈，以帮助改进 Karmada 并培育一个充满活力的协作生态系统。</p>
<p>成为 Karmada 用户组成员具有以下几个优势：</p>
<ul>
<li class=""><strong>社区认可</strong>：作为云原生多集群管理领域的领导者来展示您的组织，在 CNCF 和 Karmada 社区中获得知名度。</li>
<li class=""><strong>促进协作</strong>：与其他采用者建立联系，分享最佳实践，并在实际用例和解决方案上进行协作。</li>
<li class=""><strong>保持信息同步</strong>：及时接收重要更新通知，包括关键功能、错误修复和安全建议。</li>
<li class=""><strong>活动参与</strong>：受邀参与 Karmada 相关活动，包括 KubeCon + CloudNativeCon、网络研讨会和聚会。</li>
<li class=""><strong>职位发布</strong>：有机会在 Karmada 社区支持的职位公告板上发布与 Karmada 相关的职位空缺（目前暂不可用）。</li>
<li class=""><strong>扩展商业机会</strong>：与 Karmada 生态系统的其他成员建立潜在的商业联系和合作。</li>
</ul>
<p>在 <a href="https://github.com/karmada-io/community/tree/main/adopter-group" target="_blank" rel="noopener noreferrer" class="">GitHub 社区仓库</a> 中了解更多关于 Karmada 用户组的信息，并在 <a href="https://karmada.io/zh/adopters" target="_blank" rel="noopener noreferrer" class="">karmada.io/adopters</a> 查看完整的<code>公开</code>的采用者列表。</p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="加入-karmada-用户组">加入 Karmada 用户组<a href="https://karmada.io/zh/blog/2025/11/18/gongjikeji-officially-joined-the-karmada-adopter-group/gongjikeji#%E5%8A%A0%E5%85%A5-karmada-%E7%94%A8%E6%88%B7%E7%BB%84" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p><strong>Karmada 用户组</strong>对当前正在<strong>生产环境</strong>中使用 Karmada 的<strong>最终用户</strong>和<strong>供应商</strong>开放。这包括：</p>
<ul>
<li class=""><strong>最终用户</strong>：在其生产环境中运行 Karmada 的组织。</li>
<li class=""><strong>供应商</strong>：提供基于 Karmada 的产品或服务，并有客户在生产环境中使用这些产品或服务的公司。</li>
</ul>
<p>您是否在生产环境中使用 Karmada 并有兴趣加入 Karmada 用户组？</p>
<p>加入很简单——只需填写 <a href="https://github.com/karmada-io/karmada/issues/4540" target="_blank" rel="noopener noreferrer" class="">Karmada 用户组申请表单</a> 并提供您组织的信息。在您的申请经过审核和批准后，您的组织将被添加到 Karmada 用户组。</p>
<p>更多详情，请查看社区仓库中的 <a href="https://github.com/karmada-io/community/tree/main/adopter-group#how-to-join" target="_blank" rel="noopener noreferrer" class="">如何加入</a> 部分。</p>
<p>让我们一起发展 Karmada 社区和更广泛的多集群生态系统！</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[欢迎 Scatter Lab 加入 Karmada 用户组]]></title>
            <link>https://karmada.io/zh/blog/2025/10/16/scatterlab-officially-joined-the-karmada-adopter-group/scatterlab</link>
            <guid>https://karmada.io/zh/blog/2025/10/16/scatterlab-officially-joined-the-karmada-adopter-group/scatterlab</guid>
            <pubDate>Thu, 16 Oct 2025 00:00:00 GMT</pubDate>
            <description><![CDATA[Karmada 非常高兴地宣布 Scatter Lab 正式加入 Karmada 用户组，成为社区的重要成员。作为云原生计算基金会（CNCF）旗下的项目，Karmada 致力于为用户提供强大的多集群管理和调度能力，帮助企业在复杂的分布式环境中实现高效的应用部署和管理。Scatter Lab的加入将进一步加强 Karmada 社区，为项目的持续创新带来新的活力，标志着我们社区发展和 Karmada 在多样化生产环境中采用的又一个重要里程碑。]]></description>
            <content:encoded><![CDATA[<p>Karmada 非常高兴地宣布 <strong><a href="https://www.scatterlab.co.kr/ko" target="_blank" rel="noopener noreferrer" class="">Scatter Lab</a></strong> 正式加入 Karmada 用户组，成为社区的重要成员。作为云原生计算基金会（CNCF）旗下的项目，Karmada 致力于为用户提供强大的多集群管理和调度能力，帮助企业在复杂的分布式环境中实现高效的应用部署和管理。<strong>Scatter Lab</strong>的加入将进一步加强 Karmada 社区，为项目的持续创新带来新的活力，标志着我们社区发展和 Karmada 在多样化生产环境中采用的又一个重要里程碑。</p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="关于-scatter-lab">关于 Scatter Lab<a href="https://karmada.io/zh/blog/2025/10/16/scatterlab-officially-joined-the-karmada-adopter-group/scatterlab#%E5%85%B3%E4%BA%8E-scatter-lab" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p><img decoding="async" loading="lazy" alt="member logo" src="https://karmada.io/zh/assets/images/scatterlab-79e9d670ff2544fc88803018298248b4.png" width="6921" height="970" class="img_ev3q"></p>
<p>Scatter Lab 是一家致力于通过尖端自然语言处理和生成式人工智能技术重塑娱乐体验的韩国科技公司。公司目前以 Zeta 为核心产品，这是一款沉浸式 AI 聊天娱乐平台，于 2024 年 4 月正式上线。</p>
<p>Zeta 通过让用户与 AI 角色共同创作动态、个性化的叙事内容，重新定义了用户参与方式——从被动消费内容转变为实时、主动的共创体验。上线一年内，Zeta 注册用户已突破 200 万，月活跃用户达 80 万，用户日均使用时长超过 2 小时 40 分钟。</p>
<p>Zeta 的核心技术是 Scatter Lab 自主研发的 AI 模型 Spotwrite-1。该模型不仅追求语言连贯性，更专注于“趣味性”，通过创造力、不可预测性，甚至可控的“幻觉”机制，为用户提供引人入胜的互动体验。公司已实现从数据、训练到应用的全栈垂直整合，兼顾高性能与成本效率。</p>
<p>值得一提的是，Scatter Lab 自 2024 年第四季度起已连续六个季度实现盈利，在 AI 消费领域展现出罕见的快速增长与可持续商业化能力。</p>
<p>秉持“让 AI 成为人类叙事中的真正伙伴”这一愿景，Scatter Lab 正以 Zeta 为起点，向全球市场拓展，首站为日本，致力于塑造下一代“AI 原生娱乐”的未来。</p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="关于-karmada-用户组">关于 Karmada 用户组<a href="https://karmada.io/zh/blog/2025/10/16/scatterlab-officially-joined-the-karmada-adopter-group/scatterlab#%E5%85%B3%E4%BA%8E-karmada-%E7%94%A8%E6%88%B7%E7%BB%84" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p><strong>Karmada 用户组</strong>是一个由在其环境中成功采用 Karmada 的组织和用户组成的社区。成员分享他们的经验、最佳实践和反馈，以帮助改进 Karmada 并培育一个充满活力的协作生态系统。</p>
<p>成为 Karmada 用户组成员具有以下几个优势：</p>
<ul>
<li class=""><strong>社区认可</strong>：作为云原生多集群管理领域的领导者来展示您的组织，在 CNCF 和 Karmada 社区中获得知名度。</li>
<li class=""><strong>促进协作</strong>：与其他采用者建立联系，分享最佳实践，并在实际用例和解决方案上进行协作。</li>
<li class=""><strong>保持信息同步</strong>：及时接收重要更新通知，包括关键功能、错误修复和安全建议。</li>
<li class=""><strong>技术影响力构建</strong>：受邀参与 Karmada 相关活动，包括 KubeCon + CloudNativeCon、网络研讨会和聚会。</li>
<li class=""><strong>职位发布</strong>：有机会在 Karmada 社区支持的职位公告板上发布与 Karmada 相关的职位空缺（目前暂不可用）。</li>
<li class=""><strong>扩展商业机会</strong>：与 Karmada 生态系统的其他成员建立潜在的商业联系和合作。</li>
</ul>
<p>在 <a href="https://github.com/karmada-io/community/tree/main/adopter-group" target="_blank" rel="noopener noreferrer" class="">GitHub 社区仓库</a> 中了解更多关于 Karmada 用户组的信息，并在 <a href="https://karmada.io/zh/adopters" target="_blank" rel="noopener noreferrer" class="">karmada.io/adopters</a> 查看完整的<code>公开</code>的采用者列表。</p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="加入-karmada-用户组">加入 Karmada 用户组<a href="https://karmada.io/zh/blog/2025/10/16/scatterlab-officially-joined-the-karmada-adopter-group/scatterlab#%E5%8A%A0%E5%85%A5-karmada-%E7%94%A8%E6%88%B7%E7%BB%84" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p><strong>Karmada 用户组</strong>对当前正在<strong>生产环境</strong>中使用 Karmada 的<strong>最终用户</strong>和<strong>供应商</strong>开放。这包括：</p>
<ul>
<li class=""><strong>最终用户</strong>：在其生产环境中运行 Karmada 的组织。</li>
<li class=""><strong>供应商</strong>：提供基于 Karmada 的产品或服务，并有客户在生产环境中使用这些产品或服务的公司。</li>
</ul>
<p>您是否在生产环境中使用 Karmada 并有兴趣加入 Karmada 用户组？</p>
<p>加入很简单——只需填写 <a href="https://github.com/karmada-io/karmada/issues/4540" target="_blank" rel="noopener noreferrer" class="">Karmada 用户组申请表单</a> 并提供您组织的信息。在您的申请经过审核和批准后，您的组织将被添加到 Karmada 用户组。</p>
<p>更多详情，请查看社区仓库中的 <a href="https://github.com/karmada-io/community/tree/main/adopter-group#how-to-join" target="_blank" rel="noopener noreferrer" class="">如何加入</a> 部分。</p>
<p>让我们一起发展 Karmada 社区和更广泛的多集群生态系统！</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[欢迎彭博加入 Karmada 用户组]]></title>
            <link>https://karmada.io/zh/blog/2025/09/11/bloomberg-officially-joined-the-karmada-adopter-group/bloomberg</link>
            <guid>https://karmada.io/zh/blog/2025/09/11/bloomberg-officially-joined-the-karmada-adopter-group/bloomberg</guid>
            <pubDate>Thu, 11 Sep 2025 00:00:00 GMT</pubDate>
            <description><![CDATA[Karmada 非常高兴地宣布 彭博 正式加入 Karmada 用户组，成为社区的重要成员。作为云原生计算基金会（CNCF）旗下的项目，Karmada 致力于为用户提供强大的多集群管理和调度能力，帮助企业在复杂的分布式环境中实现高效的应用部署和管理。彭博的加入将进一步加强 Karmada 社区，为项目的持续创新带来新的活力，标志着我们社区发展和 Karmada 在多样化生产环境中采用的又一个重要里程碑。]]></description>
            <content:encoded><![CDATA[<p>Karmada 非常高兴地宣布 <strong><a href="https://www.techatbloomberg.com/opensource/" target="_blank" rel="noopener noreferrer" class="">彭博</a></strong> 正式加入 Karmada 用户组，成为社区的重要成员。作为云原生计算基金会（CNCF）旗下的项目，Karmada 致力于为用户提供强大的多集群管理和调度能力，帮助企业在复杂的分布式环境中实现高效的应用部署和管理。<strong>彭博</strong>的加入将进一步加强 Karmada 社区，为项目的持续创新带来新的活力，标志着我们社区发展和 Karmada 在多样化生产环境中采用的又一个重要里程碑。</p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="关于彭博">关于彭博<a href="https://karmada.io/zh/blog/2025/09/11/bloomberg-officially-joined-the-karmada-adopter-group/bloomberg#%E5%85%B3%E4%BA%8E%E5%BD%AD%E5%8D%9A" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p><img decoding="async" loading="lazy" alt="member logo" src="https://karmada.io/zh/assets/images/bloomberg-501467be5c8993bfb8a5c99eb0c2cafa.png" width="1320" height="598" class="img_ev3q"></p>
<p>彭博（Bloomberg）是全球商业与金融信息领域的领军者，提供值得信赖的数据、新闻与深度洞察，为市场注入透明度、提升效率并维护公平性。彭博借助可靠的技术解决方案，助力全球金融生态系统中具有影响力的各方建立连接，帮助客户做出更明智的决策，并促进更高效的协作。</p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="关于-karmada-用户组">关于 Karmada 用户组<a href="https://karmada.io/zh/blog/2025/09/11/bloomberg-officially-joined-the-karmada-adopter-group/bloomberg#%E5%85%B3%E4%BA%8E-karmada-%E7%94%A8%E6%88%B7%E7%BB%84" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p><strong>Karmada 用户组</strong>是一个由在其环境中成功采用 Karmada 的组织和用户组成的社区。成员分享他们的经验、最佳实践和反馈，以帮助改进 Karmada 并培育一个充满活力的协作生态系统。</p>
<p>成为 Karmada 用户组成员具有以下几个优势：</p>
<ul>
<li class=""><strong>社区认可</strong>：作为云原生多集群管理领域的领导者来展示您的组织，在 CNCF 和 Karmada 社区中获得知名度。</li>
<li class=""><strong>促进协作</strong>：与其他采用者建立联系，分享最佳实践，并在实际用例和解决方案上进行协作。</li>
<li class=""><strong>保持信息同步</strong>：及时接收重要更新通知，包括关键功能、错误修复和安全建议。</li>
<li class=""><strong>技术影响力构建</strong>：受邀参与 Karmada 相关活动，包括 KubeCon + CloudNativeCon、网络研讨会和聚会。</li>
<li class=""><strong>职位发布</strong>：有机会在 Karmada 社区支持的职位公告板上发布与 Karmada 相关的职位空缺（目前暂不可用）。</li>
<li class=""><strong>扩展商业机会</strong>：与 Karmada 生态系统的其他成员建立潜在的商业联系和合作。</li>
</ul>
<p>在 <a href="https://github.com/karmada-io/community/tree/main/adopter-group" target="_blank" rel="noopener noreferrer" class="">GitHub 社区仓库</a> 中了解更多关于 Karmada 用户组的信息，并在 <a href="https://karmada.io/zh/adopters" target="_blank" rel="noopener noreferrer" class="">karmada.io/adopters</a> 查看完整的<code>公开</code>的采用者列表。</p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="加入-karmada-用户组">加入 Karmada 用户组<a href="https://karmada.io/zh/blog/2025/09/11/bloomberg-officially-joined-the-karmada-adopter-group/bloomberg#%E5%8A%A0%E5%85%A5-karmada-%E7%94%A8%E6%88%B7%E7%BB%84" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p><strong>Karmada 用户组</strong>对当前正在<strong>生产环境</strong>中使用 Karmada 的<strong>最终用户</strong>和<strong>供应商</strong>开放。这包括：</p>
<ul>
<li class=""><strong>最终用户</strong>：在其生产环境中运行 Karmada 的组织。</li>
<li class=""><strong>供应商</strong>：提供基于 Karmada 的产品或服务，并有客户在生产环境中使用这些产品或服务的公司。</li>
</ul>
<p>您是否在生产环境中使用 Karmada 并有兴趣加入 Karmada 用户组？<br>
<!-- -->加入很简单——只需填写 <a href="https://github.com/karmada-io/karmada/issues/4540" target="_blank" rel="noopener noreferrer" class="">Karmada 用户组申请表单</a> 并提供您组织的信息。
在您的申请经过审核和批准后，您的组织将被添加到 Karmada 用户组。</p>
<p>更多详情，请查看社区仓库中的 <a href="https://github.com/karmada-io/community/tree/main/adopter-group#how-to-join" target="_blank" rel="noopener noreferrer" class="">如何加入</a> 部分。</p>
<p>让我们一起发展 Karmada 社区和更广泛的多集群生态系统！</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Karmada v1.15 版本发布！多模板工作负载资源感知能力增强！]]></title>
            <link>https://karmada.io/zh/blog/2025/09/05/karmada-v1.15/karmada-v1.15</link>
            <guid>https://karmada.io/zh/blog/2025/09/05/karmada-v1.15/karmada-v1.15</guid>
            <pubDate>Fri, 05 Sep 2025 00:00:00 GMT</pubDate>
            <description><![CDATA[Karmada 是开放的多云多集群容器编排引擎，旨在帮助用户在多云环境下部署和运维业务应用。凭借兼容 Kubernetes 原生 API 的能力，Karmada 可以平滑迁移单集群工作负载，并且仍可保持与 Kubernetes 周边生态工具链协同。]]></description>
            <content:encoded><![CDATA[<p>Karmada 是开放的多云多集群容器编排引擎，旨在帮助用户在多云环境下部署和运维业务应用。凭借兼容 Kubernetes 原生 API 的能力，Karmada 可以平滑迁移单集群工作负载，并且仍可保持与 Kubernetes 周边生态工具链协同。</p>
<p><a href="https://github.com/karmada-io/karmada/releases/tag/v1.15.0" target="_blank" rel="noopener noreferrer" class="">Karmada v1.15</a> 版本现已发布，本版本包含下列新增特性：</p>
<ul>
<li class="">多模板工作负载的资源精确感知</li>
<li class="">集群级故障迁移功能增强</li>
<li class="">结构化日志</li>
<li class="">Karmada 控制器和调度器性能显著提升</li>
</ul>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="新特性概览">新特性概览<a href="https://karmada.io/zh/blog/2025/09/05/karmada-v1.15/karmada-v1.15#%E6%96%B0%E7%89%B9%E6%80%A7%E6%A6%82%E8%A7%88" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<h3 class="anchor anchorTargetStickyNavbar_Vzrq" id="多模板工作负载的资源精确感知">多模板工作负载的资源精确感知<a href="https://karmada.io/zh/blog/2025/09/05/karmada-v1.15/karmada-v1.15#%E5%A4%9A%E6%A8%A1%E6%9D%BF%E5%B7%A5%E4%BD%9C%E8%B4%9F%E8%BD%BD%E7%9A%84%E8%B5%84%E6%BA%90%E7%B2%BE%E7%A1%AE%E6%84%9F%E7%9F%A5" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h3>
<p>Karmada 利用资源解释器获取工作负载的副本数和资源请求，并据此计算工作负载所需资源总量，从而实现资源感知调度，联邦配额管理等高阶能力。这种机制在传统的单模板工作负载中表现良好。然而，许多AI大数据应用的工作负载 CRD（如 FlinkDeployments，PyTorchJob 和 RayJob 等）包含多个 Pod 模板或组件，每个组件都有独特的资源需求。由于资源解释器仅能处理单个模板的资源请求，无法准确反映不同模板间的差异，导致多模板工作负载的资源计算不够精确。</p>
<p>在这个版本中，Karmada 强化了对多模板工作负载的资源感知能力，通过扩展资源解释器，Karmada 现在可以获取同一工作负载不同模板的副本数和资源请求，确保数据的精确性。这一改进也为多模板工作负载的联邦配额管理提供了更加可靠和精细的数据支持。</p>
<p>假设你部署了一个 FlinkDeployment，其资源相关配置如下：</p>
<div class="language-yaml codeBlockContainer_Ckt0 theme-code-block" style="--prism-color:#393A34;--prism-background-color:#f6f8fa"><div class="codeBlockContent_QJqH"><pre tabindex="0" class="prism-code language-yaml codeBlock_bY9V thin-scrollbar" style="color:#393A34;background-color:#f6f8fa"><code class="codeBlockLines_e6Vv"><span class="token-line" style="color:#393A34"><span class="token key atrule" style="color:#00a4db">spec</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">jobManager</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">replicas</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token number" style="color:#36acaa">1</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">resource</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">cpu</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token number" style="color:#36acaa">1</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">memory</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> 1024m</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">taskManager</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">replicas</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token number" style="color:#36acaa">1</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">resource</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">cpu</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token number" style="color:#36acaa">2</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">memory</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> 2048m</span><br></span></code></pre></div></div>
<p>通过 ResourceBinding，你可以查看资源解释器解析出的 FlinkDeployment 各个模板的副本数以及资源请求。</p>
<div class="language-yaml codeBlockContainer_Ckt0 theme-code-block" style="--prism-color:#393A34;--prism-background-color:#f6f8fa"><div class="codeBlockContent_QJqH"><pre tabindex="0" class="prism-code language-yaml codeBlock_bY9V thin-scrollbar" style="color:#393A34;background-color:#f6f8fa"><code class="codeBlockLines_e6Vv"><span class="token-line" style="color:#393A34"><span class="token key atrule" style="color:#00a4db">spec</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">components</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">name</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> jobmanager</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">replicaRequirements</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">resourceRequest</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">        </span><span class="token key atrule" style="color:#00a4db">cpu</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token string" style="color:#e3116c">"1"</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">        </span><span class="token key atrule" style="color:#00a4db">memory</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token string" style="color:#e3116c">"1.024"</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">replicas</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token number" style="color:#36acaa">1</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">name</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> taskmanager</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">replicaRequirements</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">resourceRequest</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">        </span><span class="token key atrule" style="color:#00a4db">cpu</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token string" style="color:#e3116c">"2"</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">        </span><span class="token key atrule" style="color:#00a4db">memory</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token string" style="color:#e3116c">"2.048"</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">replicas</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token number" style="color:#36acaa">1</span><br></span></code></pre></div></div>
<p>此时，FederatedResourceQuota 计算的 FlinkDeployment 占用的资源量为：</p>
<div class="language-yaml codeBlockContainer_Ckt0 theme-code-block" style="--prism-color:#393A34;--prism-background-color:#f6f8fa"><div class="codeBlockContent_QJqH"><pre tabindex="0" class="prism-code language-yaml codeBlock_bY9V thin-scrollbar" style="color:#393A34;background-color:#f6f8fa"><code class="codeBlockLines_e6Vv"><span class="token-line" style="color:#393A34"><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">status</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">overallUsed</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">cpu</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token string" style="color:#e3116c">"3"</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">memory</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> 3072m</span><br></span></code></pre></div></div>
<p>注意：该特性​​目前处于 Alpha 阶段，需要启用 MultiplePodTemplatesScheduling 特性开关​​才能使用。</p>
<p>随着多模板工作负载在云原生环境中的广泛应用，Karmada 致力于对其提供更强有力的支持。在接下来的版本中，我们将基于此功能进一步加强对多模板工作负载的调度支持，提供更加细粒度的资源感知调度——敬请期待更多更新！</p>
<p>更多有关此功能的资料请参考：<a href="https://github.com/karmada-io/karmada/tree/master/docs/proposals/scheduling/multi-podtemplate-support" target="_blank" rel="noopener noreferrer" class="">多Pod模板支持</a>。</p>
<h3 class="anchor anchorTargetStickyNavbar_Vzrq" id="集群级故障迁移功能增强">集群级故障迁移功能增强<a href="https://karmada.io/zh/blog/2025/09/05/karmada-v1.15/karmada-v1.15#%E9%9B%86%E7%BE%A4%E7%BA%A7%E6%95%85%E9%9A%9C%E8%BF%81%E7%A7%BB%E5%8A%9F%E8%83%BD%E5%A2%9E%E5%BC%BA" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h3>
<p>在之前的版本中，Karmada 提供了基本的集群级故障迁移能力，能够通过自定义的故障条件触发集群级别的应用迁移。为了满足有状态应用在集群故障迁移过程中保留其运行状态的需求，Karmada 在 v1.15 版本支持了集群故障迁移的应用状态中继机制。对于大数据处理应用（例如 Flink），利用此能力可以从故障前的 checkpoint 重新启动，无缝恢复到重启前的数据处理状态，从而避免数据重复处理。</p>
<p>社区在 PropagationPolicy/ClusterPropagationPolicy API 中的 <code>.spec.failover.cluster</code> 下引入了一个新的 <code>StatePreservation</code> 字段， 用于定义有状态应用在故障迁移期间保留和恢复状态数据的策略。结合此策略，当应用从一个故障集群迁移到另一个集群时，能够从原始资源配置中提取关键数据。</p>
<p>状态保留策略 <code>StatePreservation</code> 包含了一系列 <code>StatePreservationRule</code> 配置，通过 <code>JSONPath</code> 来指定需要保留的状态数据片段，并利用关联的 <code>AliasLabelName</code> 将数据传递到迁移后的集群。</p>
<p>以 Flink 应用为例，在 Flink 应用中，<code>jobID</code> 是一个唯一的标识符，用于区分和管理不同的 Flink 作业（jobs）。当集群发生故障时，Flink 应用可以利用 <code>jobID</code> 来恢复故障前作业的状态，从故障点处继续执行。具体的配置和步骤如下：</p>
<div class="language-yaml codeBlockContainer_Ckt0 theme-code-block" style="--prism-color:#393A34;--prism-background-color:#f6f8fa"><div class="codeBlockContent_QJqH"><pre tabindex="0" class="prism-code language-yaml codeBlock_bY9V thin-scrollbar" style="color:#393A34;background-color:#f6f8fa"><code class="codeBlockLines_e6Vv"><span class="token-line" style="color:#393A34"><span class="token key atrule" style="color:#00a4db">apiVersion</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> policy.karmada.io/v1alpha1</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">kind</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> PropagationPolicy</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">metadata</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">name</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> foo</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">spec</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token comment" style="color:#999988;font-style:italic">#...</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">failover</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">cluster</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">purgeMode</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> Directly</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">statePreservation</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">        </span><span class="token key atrule" style="color:#00a4db">rules</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">          </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">aliasLabelName</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> application.karmada.io/cluster</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">failover</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">jobid</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">            </span><span class="token key atrule" style="color:#00a4db">jsonPath</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token string" style="color:#e3116c">"{ .jobStatus.jobID }"</span><br></span></code></pre></div></div>
<ol>
<li class="">
<p>迁移前，Karmada 控制器将按照用户配置的路径提取 job ID。</p>
</li>
<li class="">
<p>迁移时，Karmada 控制器将提取的 job ID 以 label 的形式注入到 Flink 应用配置中，比如 <code>application.karmada.io/cluster-failover-jobid: &lt;jobID&gt;</code>。</p>
</li>
<li class="">
<p>运行在成员集群的 Kyverno 拦截 Flink 应用创建请求，并根据 <code>jobID</code> 获取该 job 的 checkpoint 数据存储路径，比如 <code>/&lt;shared-path&gt;/&lt;job-namespace&gt;/&lt;jobId&gt;/checkpoints/xxx</code>，然后配置 <code>initialSavepointPath</code> 指示从 save point 启动。</p>
</li>
<li class="">
<p>Flink 应用根据 <code>initialSavepointPath</code> 下的 checkpoint 数据启动，从而继承迁移前保存的最终状态。</p>
</li>
</ol>
<p>该能力广泛适用于能够基于某个 save point 启动的有状态应用程序，这些应用均可参考上述流程实现集群级故障迁移的状态中继。</p>
<p>注意：该特性​​目前处于 Alpha 阶段，需要启用 StatefulFailoverInjection 特性开关​​才能使用。</p>
<p><strong>功能约束</strong>：</p>
<ol>
<li class="">
<p>应用必须限定在单个集群中运行；</p>
</li>
<li class="">
<p>迁移清理策略（PurgeMode）限定为 <code>Directly</code>，即需要确保故障应用在旧集群上删除之后再在新集群中恢复应用，确保数据一致性。</p>
</li>
</ol>
<h3 class="anchor anchorTargetStickyNavbar_Vzrq" id="结构化日志">结构化日志<a href="https://karmada.io/zh/blog/2025/09/05/karmada-v1.15/karmada-v1.15#%E7%BB%93%E6%9E%84%E5%8C%96%E6%97%A5%E5%BF%97" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h3>
<p>日志是系统运行过程中记录事件、状态和行为的关键工具，广泛用于故障排查、性能监控和安全审计。Karmada 组件提供丰富的运行日志，帮助用户快速定位问题并回溯执行场景。在先前版本中，Karmada 仅支持非结构化的文本日志，难以被高效解析与查询，限制了其在现代化观测体系中的集成能力。</p>
<p>Karmada 在 1.15 版本引入了结构化日志支持，可通过 <code>--logging-format=json</code> 启动参数配置 JSON 格式输出。结构化日志示例如下：</p>
<div class="language-json codeBlockContainer_Ckt0 theme-code-block" style="--prism-color:#393A34;--prism-background-color:#f6f8fa"><div class="codeBlockContent_QJqH"><pre tabindex="0" class="prism-code language-json codeBlock_bY9V thin-scrollbar" style="color:#393A34;background-color:#f6f8fa"><code class="codeBlockLines_e6Vv"><span class="token-line" style="color:#393A34"><span class="token punctuation" style="color:#393A34">{</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token property" style="color:#36acaa">"ts"</span><span class="token operator" style="color:#393A34">:</span><span class="token string" style="color:#e3116c">"日志时间戳"</span><span class="token punctuation" style="color:#393A34">,</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token property" style="color:#36acaa">"logger"</span><span class="token operator" style="color:#393A34">:</span><span class="token string" style="color:#e3116c">"cluster_status_controller"</span><span class="token punctuation" style="color:#393A34">,</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token property" style="color:#36acaa">"level"</span><span class="token operator" style="color:#393A34">:</span><span class="token plain"> </span><span class="token string" style="color:#e3116c">"info"</span><span class="token punctuation" style="color:#393A34">,</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token property" style="color:#36acaa">"msg"</span><span class="token operator" style="color:#393A34">:</span><span class="token string" style="color:#e3116c">"Syncing cluster status"</span><span class="token punctuation" style="color:#393A34">,</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token property" style="color:#36acaa">"clusterName"</span><span class="token operator" style="color:#393A34">:</span><span class="token string" style="color:#e3116c">"member1"</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token punctuation" style="color:#393A34">}</span><br></span></code></pre></div></div>
<p>结构化日志的引入显著提升了日志的可用性与可观测性：</p>
<ul>
<li class="">
<p><strong>高效集成</strong>：可无缝对接 Elastic、Loki、Splunk 等主流日志系统，无需依赖复杂的正则表达式或日志解析器。</p>
</li>
<li class="">
<p><strong>高效查询</strong>：结构化字段支持快速检索与分析，显著提升故障排查效率。</p>
</li>
<li class="">
<p><strong>可观察性增强</strong>：关键上下文信息（如集群名、日志级别）以结构化字段呈现，便于跨组件、跨时间关联事件，实现精准问题定位。</p>
</li>
<li class="">
<p><strong>可维护性提升</strong>：结构化日志使开发者和运维人员在系统演进过程中更易于维护、解析和调整日志格式，保障日志体系的长期稳定与一致性。</p>
</li>
</ul>
<h3 class="anchor anchorTargetStickyNavbar_Vzrq" id="karmada-控制器和调度器性能显著提升">Karmada 控制器和调度器性能显著提升<a href="https://karmada.io/zh/blog/2025/09/05/karmada-v1.15/karmada-v1.15#karmada-%E6%8E%A7%E5%88%B6%E5%99%A8%E5%92%8C%E8%B0%83%E5%BA%A6%E5%99%A8%E6%80%A7%E8%83%BD%E6%98%BE%E8%91%97%E6%8F%90%E5%8D%87" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h3>
<p>在本次版本中，Karmada 性能优化团队继续致力于提升 Karmada 关键组件的性能，在控制器和调度器方面取得了显著进展。</p>
<p><strong>控制器方面</strong>，通过引入优先级队列，控制器能够在重启或切主后优先响应用户触发的资源变更，从而显著缩短服务重启和故障切换过程中的停机时间。</p>
<p>测试环境包含 5,000 个 Deployment、2,500 个 Policy 以及 5,000 个 ResourceBinding。在控制器重启且工作队列中仍有大量待处理事件的情况下，更新 Deployment 和 Policy。测试结果显示，<strong>控制器能够立即响应并优先处理这些更新事件，验证了该优化的有效性</strong>。</p>
<p>注意：该特性​​目前处于 Alpha 阶段，需要启用 ControllerPriorityQueue 特性开关才能使用。</p>
<p><strong>调度器方面</strong>，通过减少调度过程中的冗余计算，降低远程调用请求次数，Karmada 调度器的调度效率得到了显著提升。</p>
<p>测试记录了在开启精确调度组件 karmada-scheduler-estimator 情况下，调度 5,000 个 ResourceBinding 所用的时间，结果如下：</p>
<ul>
<li class="">调度器吞吐量 QPS 从约 15 提升至约 22，性能提升达 46%；</li>
<li class="">gRPC 请求次数从约 10,000 次减少至约 5,000 次，降幅达 50%。</li>
</ul>
<p>这些测试证明，在 1.15 版本中，Karmada 控制器和调度器的性能得到了极大提升。未来，我们将继续对控制器和调度器进行系统性的性能优化。</p>
<p>相关的详细测试报告，请参考 <a href="https://github.com/karmada-io/karmada/issues/6516" target="_blank" rel="noopener noreferrer" class="">[Performance] Overview of performance improvements for v1.15</a>。</p>
<h1>致谢贡献者</h1>
<p>Karmada v1.15 版本包含了来自 39 位贡献者的 269 次代码提交，在此对各位贡献者表示由衷的感谢：</p>
<table><thead><tr><th>^-^</th><th>^-^</th><th>^-^</th></tr></thead><tbody><tr><td>@abhi0324</td><td>@abhinav-1305</td><td>@Arhell</td></tr><tr><td>@Bhaumik10</td><td>@CaesarTY</td><td>@cbaenziger</td></tr><tr><td>@deefreak</td><td>@dekaihu</td><td>@devarsh10</td></tr><tr><td>@greenmoon55</td><td>@iawia002</td><td>@jabellard</td></tr><tr><td>@jennryaz</td><td>@liaolecheng</td><td>@linyao22</td></tr><tr><td>@LivingCcj</td><td>@liwang0513</td><td>@mohamedawnallah</td></tr><tr><td>@mohit-nagaraj</td><td>@mszacillo</td><td>@RainbowMango</td></tr><tr><td>@ritzdevp</td><td>@ryanwuer</td><td>@samzong</td></tr><tr><td>@seanlaii</td><td>@SunsetB612</td><td>@tessapham</td></tr><tr><td>@wangbowen1401</td><td>@warjiang</td><td>@wenhuwang</td></tr><tr><td>@whitewindmills</td><td>@whosefriendA</td><td>@XiShanYongYe-Chang</td></tr><tr><td>@zach593</td><td>@zclyne</td><td>@zhangsquared</td></tr><tr><td>@zhuyulicfc49</td><td>@zhzhuang-zju</td><td>@zzklachlan</td></tr></tbody></table>
<p><img decoding="async" loading="lazy" alt="karmada v1.15 contributors" src="https://karmada.io/zh/assets/images/contributors-bc4745cc1496f15d0f385242b7e576e2.png" width="601" height="750" class="img_ev3q"></p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[欢迎阿里云加入 Karmada 用户组]]></title>
            <link>https://karmada.io/zh/blog/2025/09/02/alibaba-cloud-officially-joined-the-karmada-adopter-group/alibaba-cloud</link>
            <guid>https://karmada.io/zh/blog/2025/09/02/alibaba-cloud-officially-joined-the-karmada-adopter-group/alibaba-cloud</guid>
            <pubDate>Tue, 02 Sep 2025 00:00:00 GMT</pubDate>
            <description><![CDATA[Karmada 非常高兴地宣布 阿里云 正式加入 Karmada 用户组，成为社区的重要成员。作为云原生计算基金会（CNCF）旗下的项目，Karmada 致力于为用户提供强大的多集群管理和调度能力，帮助企业在复杂的分布式环境中实现高效的应用部署和管理。阿里云的加入将进一步加强 Karmada 社区，为项目的持续创新带来新的活力，标志着我们社区发展和 Karmada 在多样化生产环境中采用的又一个重要里程碑。]]></description>
            <content:encoded><![CDATA[<p>Karmada 非常高兴地宣布 <strong><a href="https://www.alibabacloud.com/" target="_blank" rel="noopener noreferrer" class="">阿里云</a></strong> 正式加入 Karmada 用户组，成为社区的重要成员。作为云原生计算基金会（CNCF）旗下的项目，Karmada 致力于为用户提供强大的多集群管理和调度能力，帮助企业在复杂的分布式环境中实现高效的应用部署和管理。<strong>阿里云</strong>的加入将进一步加强 Karmada 社区，为项目的持续创新带来新的活力，标志着我们社区发展和 Karmada 在多样化生产环境中采用的又一个重要里程碑。</p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="关于阿里云">关于阿里云<a href="https://karmada.io/zh/blog/2025/09/02/alibaba-cloud-officially-joined-the-karmada-adopter-group/alibaba-cloud#%E5%85%B3%E4%BA%8E%E9%98%BF%E9%87%8C%E4%BA%91" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p><img decoding="async" loading="lazy" alt="member logo" src="https://karmada.io/zh/assets/images/alibabacloud-8aaf1027e6699bcba31c845a09def83f.png" width="876" height="158" class="img_ev3q"></p>
<p>阿里云（Alibaba Cloud）创立于2009年，是阿里巴巴集团旗下的全球领先的云计算与人工智能科技公司。作为中国第一家自主研发大规模分布式计算系统“飞天”的云服务商，阿里云为全球200多个国家和地区的客户提供全面的云计算服务，涵盖弹性计算、数据库、存储、网络、安全、大数据、人工智能、物联网、边缘计算和企业应用等，服务客户包括创新型中小企业、大型企业和政府机构。阿里云致力于通过技术创新，助力客户实现数字化转型，推动产业智能化升级，成为数字时代社会发展的基石。</p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="关于-karmada-用户组">关于 Karmada 用户组<a href="https://karmada.io/zh/blog/2025/09/02/alibaba-cloud-officially-joined-the-karmada-adopter-group/alibaba-cloud#%E5%85%B3%E4%BA%8E-karmada-%E7%94%A8%E6%88%B7%E7%BB%84" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p><strong>Karmada 用户组</strong>是一个由在其环境中成功采用 Karmada 的组织和用户组成的社区。成员分享他们的经验、最佳实践和反馈，以帮助改进 Karmada 并培育一个充满活力的协作生态系统。</p>
<p>成为 Karmada 用户组成员具有以下几个优势：</p>
<ul>
<li class=""><strong>社区认可</strong>：作为云原生多集群管理领域的领导者来展示您的组织，在 CNCF 和 Karmada 社区中获得知名度。</li>
<li class=""><strong>促进协作</strong>：与其他采用者建立联系，分享最佳实践，并在实际用例和解决方案上进行协作。</li>
<li class=""><strong>保持信息同步</strong>：及时接收重要更新通知，包括关键功能、错误修复和安全建议。</li>
<li class=""><strong>技术影响力构建</strong>：受邀参与 Karmada 相关活动，包括 KubeCon + CloudNativeCon、网络研讨会和聚会。</li>
<li class=""><strong>职位发布</strong>：有机会在 Karmada 社区支持的职位公告板上发布与 Karmada 相关的职位空缺（目前暂不可用）。</li>
<li class=""><strong>扩展商业机会</strong>：与 Karmada 生态系统的其他成员建立潜在的商业联系和合作。</li>
</ul>
<p>在 <a href="https://github.com/karmada-io/community/tree/main/adopter-group" target="_blank" rel="noopener noreferrer" class="">GitHub 社区仓库</a> 中了解更多关于 Karmada 用户组的信息，并在 <a href="https://karmada.io/zh/adopters" target="_blank" rel="noopener noreferrer" class="">karmada.io/adopters</a> 查看完整的<code>公开</code>的采用者列表。</p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="加入-karmada-用户组">加入 Karmada 用户组<a href="https://karmada.io/zh/blog/2025/09/02/alibaba-cloud-officially-joined-the-karmada-adopter-group/alibaba-cloud#%E5%8A%A0%E5%85%A5-karmada-%E7%94%A8%E6%88%B7%E7%BB%84" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p><strong>Karmada 用户组</strong>对当前正在<strong>生产环境</strong>中使用 Karmada 的<strong>最终用户</strong>和<strong>供应商</strong>开放。这包括：</p>
<ul>
<li class=""><strong>最终用户</strong>：在其生产环境中运行 Karmada 的组织。</li>
<li class=""><strong>供应商</strong>：提供基于 Karmada 的产品或服务，并有客户在生产环境中使用这些产品或服务的公司。</li>
</ul>
<p>您是否在生产环境中使用 Karmada 并有兴趣加入 Karmada 用户组？<br>
<!-- -->加入很简单——只需填写 <a href="https://github.com/karmada-io/karmada/issues/4540" target="_blank" rel="noopener noreferrer" class="">Karmada 用户组申请表单</a> 并提供您组织的信息。
在您的申请经过审核和批准后，您的组织将被添加到 Karmada 用户组。</p>
<p>更多详情，请查看社区仓库中的 <a href="https://github.com/karmada-io/community/tree/main/adopter-group#how-to-join" target="_blank" rel="noopener noreferrer" class="">如何加入</a> 部分。</p>
<p>让我们一起发展 Karmada 社区和更广泛的多集群生态系统！</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[欢迎中科类脑加入 Karmada 用户组]]></title>
            <link>https://karmada.io/zh/blog/2025/07/30/brain-inspired-officially-joined-the-karmada-adopter-group/brain-inspired</link>
            <guid>https://karmada.io/zh/blog/2025/07/30/brain-inspired-officially-joined-the-karmada-adopter-group/brain-inspired</guid>
            <pubDate>Wed, 30 Jul 2025 00:00:00 GMT</pubDate>
            <description><![CDATA[Karmada 非常高兴地宣布 中科类脑 正式加入 Karmada 用户组，成为社区的重要成员。作为云原生计算基金会（CNCF）旗下的项目，Karmada 致力于为用户提供强大的多集群管理和调度能力，帮助企业在复杂的分布式环境中实现高效的应用部署和管理。中科类脑的加入将进一步加强 Karmada 社区，为项目的持续创新带来新的活力，标志着我们社区发展和 Karmada 在多样化生产环境中采用的又一个重要里程碑。]]></description>
            <content:encoded><![CDATA[<p>Karmada 非常高兴地宣布 <strong><a href="http://www.leinao.ai/" target="_blank" rel="noopener noreferrer" class="">中科类脑</a></strong> 正式加入 Karmada 用户组，成为社区的重要成员。作为云原生计算基金会（CNCF）旗下的项目，Karmada 致力于为用户提供强大的多集群管理和调度能力，帮助企业在复杂的分布式环境中实现高效的应用部署和管理。<strong>中科类脑</strong>的加入将进一步加强 Karmada 社区，为项目的持续创新带来新的活力，标志着我们社区发展和 Karmada 在多样化生产环境中采用的又一个重要里程碑。</p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="关于中科类脑">关于中科类脑<a href="https://karmada.io/zh/blog/2025/07/30/brain-inspired-officially-joined-the-karmada-adopter-group/brain-inspired#%E5%85%B3%E4%BA%8E%E4%B8%AD%E7%A7%91%E7%B1%BB%E8%84%91" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p><img decoding="async" loading="lazy" alt="brain inspired" src="data:image/png;base64,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" width="239" height="95" class="img_ev3q"></p>
<p>合肥中科类脑智能技术有限公司成立于2017年，是一家专注于类脑智能技术研发与应用的国家高新技术企业、国家级专精特新“小巨人”企业。公司在机器视觉大模型、小样本学习、因果视觉与因果推理、稳定学习、类脑博弈优化决策等多个人工智能前沿技术领域处于行业先进地位，广泛应用于算力基础设施、智慧能源和算电碳协同发展三大业务领域。</p>
<p>中科类脑秉承“推动前沿智能技术落地，助力产业数智升级”的使命，持续推出创新的智能化产品及解决方案，力求打造垂直领域人工智能应用的深度闭环。公司致力于成为全球领先的能源智能服务企业，致力于成为全球AI生态建设者。</p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="关于-karmada-用户组">关于 Karmada 用户组<a href="https://karmada.io/zh/blog/2025/07/30/brain-inspired-officially-joined-the-karmada-adopter-group/brain-inspired#%E5%85%B3%E4%BA%8E-karmada-%E7%94%A8%E6%88%B7%E7%BB%84" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p><strong>Karmada 用户组</strong>是一个由在其环境中成功采用 Karmada 的组织和用户组成的社区。成员分享他们的经验、最佳实践和反馈，以帮助改进 Karmada 并培育一个充满活力的协作生态系统。</p>
<p>成为 Karmada 用户组成员具有以下几个优势：</p>
<ul>
<li class=""><strong>社区认可</strong>：作为云原生多集群管理领域的领导者来展示您的组织，在 CNCF 和 Karmada 社区中获得知名度。</li>
<li class=""><strong>促进协作</strong>：与其他采用者建立联系，分享最佳实践，并在实际用例和解决方案上进行协作。</li>
<li class=""><strong>保持信息同步</strong>：及时接收重要更新通知，包括关键功能、错误修复和安全建议。</li>
<li class=""><strong>技术影响力构建</strong>：受邀参与 Karmada 相关活动，包括 KubeCon + CloudNativeCon、网络研讨会和聚会。</li>
<li class=""><strong>职位发布</strong>：有机会在 Karmada 社区支持的职位公告板上发布与 Karmada 相关的职位空缺（目前暂不可用）。</li>
<li class=""><strong>扩展商业机会</strong>：与 Karmada 生态系统的其他成员建立潜在的商业联系和合作。</li>
</ul>
<p>在 <a href="https://github.com/karmada-io/community/tree/main/adopter-group" target="_blank" rel="noopener noreferrer" class="">GitHub 社区仓库</a> 中了解更多关于 Karmada 用户组的信息，并在 <a href="https://karmada.io/zh/adopters" target="_blank" rel="noopener noreferrer" class="">karmada.io/adopters</a> 查看完整的<code>公开</code>的采用者列表。</p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="加入-karmada-用户组">加入 Karmada 用户组<a href="https://karmada.io/zh/blog/2025/07/30/brain-inspired-officially-joined-the-karmada-adopter-group/brain-inspired#%E5%8A%A0%E5%85%A5-karmada-%E7%94%A8%E6%88%B7%E7%BB%84" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p><strong>Karmada 用户组</strong>对当前正在<strong>生产环境</strong>中使用 Karmada 的<strong>最终用户</strong>和<strong>供应商</strong>开放。这包括：</p>
<ul>
<li class=""><strong>最终用户</strong>：在其生产环境中运行 Karmada 的组织。</li>
<li class=""><strong>供应商</strong>：提供基于 Karmada 的产品或服务，并有客户在生产环境中使用这些产品或服务的公司。</li>
</ul>
<p>您是否在生产环境中使用 Karmada 并有兴趣加入 Karmada 用户组？<br>
<!-- -->加入很简单——只需填写 <a href="https://github.com/karmada-io/karmada/issues/4540" target="_blank" rel="noopener noreferrer" class="">Karmada 用户组申请表单</a> 并提供您组织的信息。
在您的申请经过审核和批准后，您的组织将被添加到 Karmada 用户组。</p>
<p>更多详情，请查看社区仓库中的 <a href="https://github.com/karmada-io/community/tree/main/adopter-group#how-to-join" target="_blank" rel="noopener noreferrer" class="">如何加入</a> 部分。</p>
<p>让我们一起发展 Karmada 社区和更广泛的多集群生态系统！</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Karmada v1.14 版本发布！新增联邦资源配额管理能力!]]></title>
            <link>https://karmada.io/zh/blog/2025/05/30/karmada-v1.14/karmada-v1.14</link>
            <guid>https://karmada.io/zh/blog/2025/05/30/karmada-v1.14/karmada-v1.14</guid>
            <pubDate>Fri, 30 May 2025 00:00:00 GMT</pubDate>
            <description><![CDATA[Karmada 是开放的多云多集群容器编排引擎，旨在帮助用户在多云环境下部署和运维业务应用。凭借兼容 Kubernetes 原生 API 的能力，Karmada 可以平滑迁移单集群工作负载，并且仍可保持与 Kubernetes 周边生态工具链协同。]]></description>
            <content:encoded><![CDATA[<p>Karmada 是开放的多云多集群容器编排引擎，旨在帮助用户在多云环境下部署和运维业务应用。凭借兼容 Kubernetes 原生 API 的能力，Karmada 可以平滑迁移单集群工作负载，并且仍可保持与 Kubernetes 周边生态工具链协同。</p>
<p><a href="https://github.com/karmada-io/karmada/releases/tag/v1.14.0" target="_blank" rel="noopener noreferrer" class="">Karmada v1.14</a> 版本现已发布，本版本包含下列新增特性：</p>
<ul>
<li class="">新增联邦资源配额管理能力，用于多租户场景下资源治理</li>
<li class="">新增定制化污点管理能力，消除隐式集群故障迁移</li>
<li class="">Karmada Operator 功能持续演进</li>
<li class="">Karmada 控制器性能显著提升</li>
</ul>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="新特性概览">新特性概览<a href="https://karmada.io/zh/blog/2025/05/30/karmada-v1.14/karmada-v1.14#%E6%96%B0%E7%89%B9%E6%80%A7%E6%A6%82%E8%A7%88" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<h3 class="anchor anchorTargetStickyNavbar_Vzrq" id="联邦资源配额管理">联邦资源配额管理<a href="https://karmada.io/zh/blog/2025/05/30/karmada-v1.14/karmada-v1.14#%E8%81%94%E9%82%A6%E8%B5%84%E6%BA%90%E9%85%8D%E9%A2%9D%E7%AE%A1%E7%90%86" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h3>
<p>在多租户的云基础设施中，配额管理是确保资源公平分配和防止超额使用的关键。尤其在多云多集群环境下，分散的配额系统往往导致资源监控困难和管理割裂，因此实现跨集群的联邦配额管理成为提升资源治理效率的核心要素。</p>
<p>此前，Karmada 通过 FederatedResourceQuota 将全局配额分配至成员集群，由各集群本地实施配额管控。本次版本升级增强了联邦配额管理能力，新增控制平面全局配额检查机制，支持直接在控制平面进行全局资源配额校验。</p>
<p>该功能特别适用于以下场景：</p>
<ul>
<li class="">您需要从统一位置跟踪资源消耗和限制，而无需关注集群级别的分配情况。</li>
<li class="">您希望通过验证配额限制来避免超额的任务提交。</li>
</ul>
<p>注意：该特性目前处于 <code>Alpha</code> 阶段，需要启用 <code>FederatedQuotaEnforcement</code> Feature Gate 才能使用。</p>
<p>假设您想设置总体 CPU 限制为 100，您可以按照如下配置进行定义：</p>
<div class="language-yaml codeBlockContainer_Ckt0 theme-code-block" style="--prism-color:#393A34;--prism-background-color:#f6f8fa"><div class="codeBlockContent_QJqH"><pre tabindex="0" class="prism-code language-yaml codeBlock_bY9V thin-scrollbar" style="color:#393A34;background-color:#f6f8fa"><code class="codeBlockLines_e6Vv"><span class="token-line" style="color:#393A34"><span class="token key atrule" style="color:#00a4db">apiVersion</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> policy.karmada.io/v1alpha1</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">kind</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> FederatedResourceQuota</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">metadata</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">name</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> team</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">foo</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">namespace</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> team</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">foo</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">spec</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">overall</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">cpu</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token number" style="color:#36acaa">100</span><br></span></code></pre></div></div>
<p>一旦应用，Karmada 将开始监控和执行 test 命名空间的 CPU 资源限制。假设您应用了一个需要 20 个 CPU 的新 Deployment。联邦资源配额的状态将更新为如下所示：</p>
<div class="language-yaml codeBlockContainer_Ckt0 theme-code-block" style="--prism-color:#393A34;--prism-background-color:#f6f8fa"><div class="codeBlockContent_QJqH"><pre tabindex="0" class="prism-code language-yaml codeBlock_bY9V thin-scrollbar" style="color:#393A34;background-color:#f6f8fa"><code class="codeBlockLines_e6Vv"><span class="token-line" style="color:#393A34"><span class="token key atrule" style="color:#00a4db">spec</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">overall</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">cpu</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token number" style="color:#36acaa">100</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">status</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">overall</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">cpu</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token number" style="color:#36acaa">100</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">overallUsed</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">cpu</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token number" style="color:#36acaa">20</span><br></span></code></pre></div></div>
<p>如果您应用的资源超过 100 个CPU的限制，该资源将不会被调度到您的成员集群。</p>
<p>有关此功能的详细用法，可以参考特性使用文档：<a href="https://karmada.io/zh/docs/userguide/bestpractices/federated-resource-quota/" target="_blank" rel="noopener noreferrer" class="">Federated ResourceQuota</a>。</p>
<h3 class="anchor anchorTargetStickyNavbar_Vzrq" id="定制化污点管理">定制化污点管理<a href="https://karmada.io/zh/blog/2025/05/30/karmada-v1.14/karmada-v1.14#%E5%AE%9A%E5%88%B6%E5%8C%96%E6%B1%A1%E7%82%B9%E7%AE%A1%E7%90%86" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h3>
<p>在 v1.14 之前的版本中，当用户启用故障转移功能时，系统在检测到健康状态异常后会自动向集群添加一个 NoExecute effect 污点，从而触发目标集群上所有资源的迁移。</p>
<p>在这个版本中，我们对系统中潜在的迁移触发因素进行了全面审查。所有隐含的集群故障转移行为已被消除，并且引入了针对集群故障机制的明确约束条件。这使得因集群故障而引发的资源迁移能够得到统一管理，进一步增强了系统的稳定性和可预测性。</p>
<p>集群故障条件是通过评估出现故障的集群对象的状态条件来确定的，以便应用污点，这一过程可以称为“Taint Cluster By Conditions”。此版本引入了一个新的 API - ClusterTaintPolicy，它允许用户自定义规则，以便在预定义的集群状态条件得到满足时，为目标集群添加特定的污点。</p>
<p><img decoding="async" loading="lazy" alt="Cluster Taint Management" src="https://karmada.io/zh/assets/images/clustertaintpolicy-a9c190fe0cfa6cd255462a95af527d4f.png" width="871" height="388" class="img_ev3q"></p>
<p>对于更复杂的集群故障判断场景，用户可以直接实现一个自定义的“集群污点控制器”，以控制如何向集群对象添加或移除污点。</p>
<p>ClusterTaintPolicy 是一种 Cluster scope 资源，下面我们给一个简单的例子来说明它的用法：</p>
<div class="language-yaml codeBlockContainer_Ckt0 theme-code-block" style="--prism-color:#393A34;--prism-background-color:#f6f8fa"><div class="codeBlockContent_QJqH"><pre tabindex="0" class="prism-code language-yaml codeBlock_bY9V thin-scrollbar" style="color:#393A34;background-color:#f6f8fa"><code class="codeBlockLines_e6Vv"><span class="token-line" style="color:#393A34"><span class="token key atrule" style="color:#00a4db">apiVersion</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> policy.karmada.io/v1alpha1</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">kind</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> ClusterTaintPolicy</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">metadata</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">name</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> detect</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">cluster</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">notready</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">spec</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">targetClusters</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">clusterNames</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> member1</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> member2</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">addOnConditions</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">conditionType</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> Ready</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">operator</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> NotIn</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">statusValues</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token string" style="color:#e3116c">"True"</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">conditionType</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> NetworkAvailable</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">operator</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> NotIn</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">statusValues</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token string" style="color:#e3116c">"True"</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">removeOnConditions</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">conditionType</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> Ready</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">operator</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> In</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">statusValues</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token string" style="color:#e3116c">"True"</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">conditionType</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> NetworkAvailable</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">operator</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> In</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">statusValues</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token string" style="color:#e3116c">"True"</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">taints</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">key</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> not</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">ready</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">effect</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> NoSchedule</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">key</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> not</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">ready</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">effect</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> NoExecute</span><br></span></code></pre></div></div>
<p>上面的例子描述了一个针对 <code>member1</code> 和 <code>member2</code> 集群的 ClusterTaintPolicy 资源，当集群的状态条件同时满足 <code>Type</code> 为 <code>Ready</code> 和 <code>NetworkAvailable</code> 的 condition value 不等于 True 时，会为目标集群添加污点 <code>{not-ready:NoSchedule}</code> 与 <code>{not-ready:NoExecute}</code>；当集群的状态条件同时满足 <code>Type</code> 为 <code>Ready</code> 和 <code>NetworkAvailable</code> 的 condition value 等于 True 时，会移除目标集群上的污点 <code>{not-ready:NoSchedule}</code> 和 <code>{not-ready:NoExecute}</code>。</p>
<p>有关此功能的详细用法，可以参考特性使用文档：<a href="https://karmada.io/docs/next/userguide/failover/cluster-taint-management/" target="_blank" rel="noopener noreferrer" class="">集群污点管理</a>。</p>
<h3 class="anchor anchorTargetStickyNavbar_Vzrq" id="karmada-operator-功能持续演进">Karmada Operator 功能持续演进<a href="https://karmada.io/zh/blog/2025/05/30/karmada-v1.14/karmada-v1.14#karmada-operator-%E5%8A%9F%E8%83%BD%E6%8C%81%E7%BB%AD%E6%BC%94%E8%BF%9B" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h3>
<p>本版本持续增强 Karmada Operator，新增以下功能：</p>
<ul>
<li class="">支持配置 Leaf 证书有效期。</li>
<li class="">支持 Karmada 控制平面暂停调谐。</li>
<li class="">支持为 <code>karmada-webhook</code> 组件配置 feature gates。</li>
<li class="">支持为 <code>karmada-apiserver</code> 组件执行 <code>loadBalancerClass</code> 以选择特定的负载均衡实现。</li>
<li class="">引入 <code>karmada_build_info</code> 指标来展示构建信息，以及一组运行时指标。</li>
</ul>
<p>这些改进使得 <code>karmada-operator</code> 更加灵活且可定制，提高了整个 Karmada 系统的可靠性和稳定性。</p>
<h3 class="anchor anchorTargetStickyNavbar_Vzrq" id="karmada-控制器性能显著提升">Karmada 控制器性能显著提升<a href="https://karmada.io/zh/blog/2025/05/30/karmada-v1.14/karmada-v1.14#karmada-%E6%8E%A7%E5%88%B6%E5%99%A8%E6%80%A7%E8%83%BD%E6%98%BE%E8%91%97%E6%8F%90%E5%8D%87" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h3>
<p>自 1.13 版本发布以来，Karmada adopters 自发组织起来对 Karmada 性能进行优化。如今，一个稳定且持续运作的性能优化团队 SIG-Scalability 已经组建，致力于提升 Karmada 的性能与稳定性。感谢所有参与者付出的努力。如果大家有兴趣，随时欢迎大家加入。</p>
<p>在本次版本中，Karmada 实现了显著的性能提升，尤其是在 <code>karmada-controller-manager</code> 组件中。为验证这些改进，实施了以下测试设置：</p>
<p>测试设置包括 5000 个 Deployment，每个 Deployment 都与一个相应的 PropagationPolicy 配对，该策略将其调度到两个成员集群。每个 Deployment 还依赖一个唯一的 ConfigMap，它会与 Deployment 一起分发到相同的集群。这些资源是在 <code>karmada-controller-manager</code> 组件离线时创建的，这
意味着在测试期间 Karmada 首次对它们进行同步。测试结果如下：</p>
<ul>
<li class="">冷启动时间（清空工作队列）从约 7 分钟缩短至约 4 分钟，提升了 45%。</li>
<li class="">资源检测器：平均处理时间的最大值从 391 毫秒降至 180 毫秒（提升了 54%）。</li>
<li class="">依赖分发器：平均处理时间的最大值从 378 毫秒降至 216 毫秒（提升了 43%）。</li>
<li class="">执行控制器：平均处理时间的最大值从 505 毫秒降至 248 毫秒（提升了 50%）。</li>
</ul>
<p>除了更快的处理速度，资源消耗也显著降低：</p>
<ul>
<li class="">CPU使用率从 4 - 7.5 核降至 1.8 - 2.4 核（降幅 40% - 65%）。</li>
<li class="">内存峰值使用量从 1.9 GB 降至 1.47 GB（降幅 22%）。</li>
</ul>
<p>这些数据证明，在 1.14 版本中，Karmada 控制器的性能得到了极大提升。未来，我们将继续对控制器和调度器进行系统性的性能优化。</p>
<p>相关的详细测试报告，请参考 <a href="https://github.com/karmada-io/karmada/issues/6394" target="_blank" rel="noopener noreferrer" class="">[Performance] Overview of performance improvements for v1.14</a>。</p>
<h1>致谢贡献者</h1>
<p>Karmada v1.14 版本包含了来自 30 位贡献者的 271 次代码提交，在此对各位贡献者表示由衷的感谢：</p>
<table><thead><tr><th>^-^</th><th>^-^</th><th>^-^</th></tr></thead><tbody><tr><td>@Arhell</td><td>@baiyutang</td><td>@chaosi-zju</td></tr><tr><td>@CharlesQQ</td><td>@dongjiang1989</td><td>@everpeace</td></tr><tr><td>@husnialhamdani</td><td>@ikaven1024</td><td>@jabellard</td></tr><tr><td>@liangyuanpeng</td><td>@likakuli</td><td>@LivingCcj</td></tr><tr><td>@liwang0513</td><td>@MdSayemkhan</td><td>@mohamedawnallah</td></tr><tr><td>@mojojoji</td><td>@mszacillo</td><td>@my-git9</td></tr><tr><td>@Pratham-B-Parlecha</td><td>@RainbowMango</td><td>@rajsinghtech</td></tr><tr><td>@seanlaii</td><td>@tangzhongren</td><td>@tiansuo114</td></tr><tr><td>@vie-serendipity</td><td>@warjiang</td><td>@whosefriendA</td></tr><tr><td>@XiShanYongYe-Chang</td><td>@zach593</td><td>@zhzhuang-zju</td></tr></tbody></table>
<p><img decoding="async" loading="lazy" alt="karmada v1.14 contributors" src="https://karmada.io/zh/assets/images/contributors-408ce724c9a68ad5cb2b75b227a574ee.png" width="593" height="512" class="img_ev3q"></p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Karmada v1.13 版本发布！新增应用优先级调度能力]]></title>
            <link>https://karmada.io/zh/blog/2025/03/19/karmada-v1.13/karmada-v1.13</link>
            <guid>https://karmada.io/zh/blog/2025/03/19/karmada-v1.13/karmada-v1.13</guid>
            <pubDate>Wed, 19 Mar 2025 00:00:00 GMT</pubDate>
            <description><![CDATA[Karmada 是开放的多云多集群容器编排引擎，旨在帮助用户在多云环境下部署和运维业务应用。凭借兼容 Kubernetes 原生 API 的能力，Karmada 可以平滑迁移单集群工作负载，并且仍可保持与 Kubernetes 周边生态工具链协同。]]></description>
            <content:encoded><![CDATA[<p>Karmada 是开放的多云多集群容器编排引擎，旨在帮助用户在多云环境下部署和运维业务应用。凭借兼容 Kubernetes 原生 API 的能力，Karmada 可以平滑迁移单集群工作负载，并且仍可保持与 Kubernetes 周边生态工具链协同。</p>
<p><img decoding="async" loading="lazy" alt="karmada" src="https://karmada.io/zh/assets/images/karmada-a74e718f9de7807cf77d2ca50a3a45cc.png" width="629" height="246" class="img_ev3q"></p>
<p><a href="https://github.com/karmada-io/karmada/releases/tag/v1.13.0" target="_blank" rel="noopener noreferrer" class="">Karmada v1.13 版本</a> 现已发布，本版本包含下列新增特性：</p>
<ul>
<li class="">新增应用优先级调度功能，可用于保证关键作业时效性</li>
<li class="">新增应用调度暂停与恢复功能，可用于构建多集群队列系统</li>
<li class="">Karmada Operator 功能持续演进</li>
<li class="">Karmada 控制器性能优化提升</li>
<li class="">Karmada Dashboard 首个版本发布！开启多云编排可视化新篇章</li>
</ul>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="新特性概览">新特性概览<a href="https://karmada.io/zh/blog/2025/03/19/karmada-v1.13/karmada-v1.13#%E6%96%B0%E7%89%B9%E6%80%A7%E6%A6%82%E8%A7%88" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<h3 class="anchor anchorTargetStickyNavbar_Vzrq" id="应用优��先级调度">应用优先级调度<a href="https://karmada.io/zh/blog/2025/03/19/karmada-v1.13/karmada-v1.13#%E5%BA%94%E7%94%A8%E4%BC%98%E5%85%88%E7%BA%A7%E8%B0%83%E5%BA%A6" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h3>
<p>当前，Karmada 调度器调度负载时遵循 FIFO 策略，按照负载进入调度队列的次序依次调度。但在某些场景，比如AI训练类作业平台，用户希望工作负载能够按照优先级进行调度。这样当高优先级、紧急任务进入调度队列时，可以实现"插队"的效果，从而确保关键业务的时效性与服务质量。</p>
<p>从 v1.13.0 版本起，用户使用 PropagationPolicy 分发负载时可以指定调度优先级(.spec.SchedulePriority字段)。SchedulePriority 会指向用户事先配置的优先级类，Karmada 解析该优先级类并获取对应的优先级数值，值越大，调度优先级越高。</p>
<p>例如，需要在环境中部署 A，B 两种应用，应用A 负责在线交易这类对实时性要求高的业务，而应用B 负责定期日志清理等非紧急且对时间不敏感的业务。为确保在资源紧张时，应用A 能优先被调度，可以给应用A 配置较大的优先级类。</p>
<p>应用A 的优先级类和 PropagationPolicy 配置：</p>
<div class="language-yaml codeBlockContainer_Ckt0 theme-code-block" style="--prism-color:#393A34;--prism-background-color:#f6f8fa"><div class="codeBlockContent_QJqH"><pre tabindex="0" class="prism-code language-yaml codeBlock_bY9V thin-scrollbar" style="color:#393A34;background-color:#f6f8fa"><code class="codeBlockLines_e6Vv"><span class="token-line" style="color:#393A34"><span class="token key atrule" style="color:#00a4db">apiVersion</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> scheduling.k8s.io/v1</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">kind</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> PriorityClass</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">metadata</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">   </span><span class="token key atrule" style="color:#00a4db">name</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> high</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">priority</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">value</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token number" style="color:#36acaa">1000000</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain" style="display:inline-block"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token punctuation" style="color:#393A34">---</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">apiVersion</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> policy.karmada.io/v1alpha1</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">kind</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> PropagationPolicy</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">metadata</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">name</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> propagation</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">a</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">spec</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">   </span><span class="token key atrule" style="color:#00a4db">schedulePriority</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">priorityClassSource</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> KubePriorityClass</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">priorityClassName</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> high</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">priority</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">   </span><span class="token key atrule" style="color:#00a4db">placement</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">clusterAffinity</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">         </span><span class="token key atrule" style="color:#00a4db">clusterNames</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">            </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> member1</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">   </span><span class="token key atrule" style="color:#00a4db">resourceSelectors</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">apiVersion</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> apps/v1</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">        </span><span class="token key atrule" style="color:#00a4db">kind</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> Deployment</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">        </span><span class="token key atrule" style="color:#00a4db">name</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> application</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">a</span><br></span></code></pre></div></div>
<p>应用B 的优先级类和 PropagationPolicy 配置：</p>
<div class="language-yaml codeBlockContainer_Ckt0 theme-code-block" style="--prism-color:#393A34;--prism-background-color:#f6f8fa"><div class="codeBlockContent_QJqH"><pre tabindex="0" class="prism-code language-yaml codeBlock_bY9V thin-scrollbar" style="color:#393A34;background-color:#f6f8fa"><code class="codeBlockLines_e6Vv"><span class="token-line" style="color:#393A34"><span class="token key atrule" style="color:#00a4db">apiVersion</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> scheduling.k8s.io/v1</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">kind</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> PriorityClass</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">metadata</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">   </span><span class="token key atrule" style="color:#00a4db">name</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> low</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">priority</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">value</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token number" style="color:#36acaa">1</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain" style="display:inline-block"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token punctuation" style="color:#393A34">---</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">apiVersion</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> policy.karmada.io/v1alpha1</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">kind</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> PropagationPolicy</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">metadata</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">name</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> propagation</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">b</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">spec</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">   </span><span class="token key atrule" style="color:#00a4db">schedulePriority</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">priorityClassSource</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> KubePriorityClass</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">priorityClassName</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> low</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">priority</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">   </span><span class="token key atrule" style="color:#00a4db">placement</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">clusterAffinity</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">         </span><span class="token key atrule" style="color:#00a4db">clusterNames</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">            </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> member1</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">   </span><span class="token key atrule" style="color:#00a4db">resourceSelectors</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">apiVersion</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> apps/v1</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">        </span><span class="token key atrule" style="color:#00a4db">kind</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> Deployment</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">        </span><span class="token key atrule" style="color:#00a4db">name</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> application</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">b</span><br></span></code></pre></div></div>
<p>通过上述配置，当调度队列同时存在应用A 和 应用B 的工作负载时，应用A 会被优先调度，即便应用B 先进入队列。</p>
<p>更多有关应用优先级调度的资料请参考：<a href="https://github.com/karmada-io/karmada/tree/master/docs/proposals/scheduling/binding-priority-preemption" target="_blank" rel="noopener noreferrer" class="">应用优先级调度</a></p>
<h3 class="anchor anchorTargetStickyNavbar_Vzrq" id="应用调度暂停与恢复">应用调度暂停与恢复<a href="https://karmada.io/zh/blog/2025/03/19/karmada-v1.13/karmada-v1.13#%E5%BA%94%E7%94%A8%E8%B0%83%E5%BA%A6%E6%9A%82%E5%81%9C%E4%B8%8E%E6%81%A2%E5%A4%8D" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h3>
<p>应用从 Karmada 控制面分发到成员集群大致可以分为两个环节：</p>
<ul>
<li class=""><strong>应用调度：</strong> Karmada 调度器根据分发策略为应用选择合适的集群。</li>
<li class=""><strong>应用分发：</strong> Karmada 控制器按照调度结果将应用分发到指定的集群。</li>
</ul>
<p>Karmada 现已支持应用调度及分发过程的启停控制，这一能力为用户提供了灵活的调度管理手段。基于此，用户可根据实际业务需求，构建定制化的高级调度策略。例如，可将其与业务发布流程相结合，实现联邦应用跨集群的滚动升级；也可结合工作负载优先级，达成队列与工作负载的优先级调度。其中应用分发环节的暂停与恢复是在 v1.11.0 版本引入的功能，详细请参考：<a href="https://karmada.io/docs/next/userguide/scheduling/resource-propagating/#suspend-and-resume-of-resource-propagation" target="_blank" rel="noopener noreferrer" class="">应用分发暂停与恢复</a>。而在 v1.13.0 版本，Karmada 正式引入了应用调度环节的暂停与恢复。</p>
<p>社区在字段 ResourceBinding.Spec.Suspension 中新增了字段 Scheduling，用于控制应用调度的暂停与恢复。当 Suspension.Scheduling 为 true 时，应用调度暂停。当 Suspension.Scheduling 为 false 时，应用调度恢复。此功能可以与第三方系统集成，通过控制 ResourceBinding 的 Suspension.Scheduling 的字段值，实现对应用调度过程的精准控制。例如，在创建 ResourceBinding 时，可通过 webhook 将 Suspension.Scheduling 设置为 true 以暂停调度；随后，可对应用进行优先级排序、容量规划等操作；待相关操作完成后，设置 Suspension.Scheduling 为 false 即可恢复调度，最终实现有序调度/资源配额管理等高级能力。</p>
<p>更多有关应用调度暂停与恢复的资料请参考：<a href="https://github.com/karmada-io/karmada/tree/master/docs/proposals/scheduling-suspension" target="_blank" rel="noopener noreferrer" class="">应用调度暂停与恢复能力</a></p>
<h3 class="anchor anchorTargetStickyNavbar_Vzrq" id="karmada-operator-功能持续演进">Karmada Operator 功能持续演进<a href="https://karmada.io/zh/blog/2025/03/19/karmada-v1.13/karmada-v1.13#karmada-operator-%E5%8A%9F%E8%83%BD%E6%8C%81%E7%BB%AD%E6%BC%94%E8%BF%9B" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h3>
<p>本版本持续增强 Karmada Operator，新增以下功能：</p>
<ul>
<li class="">
<p>支持配置 API Server 边车容器：用户可通过此功能为 Karmada API Server 配置边车容器，从而集成辅助服务。例如，可集成 <a href="https://kubernetes.io/docs/tasks/administer-cluster/kms-provider/" target="_blank" rel="noopener noreferrer" class="">KMS 插件</a>，实现静态数据加密。</p>
<p>更多有关 API Server 边车容器配置的资料请参考：<a href="https://github.com/karmada-io/karmada/tree/master/docs/proposals/karmada-operator/api-server-sidecard-containers" target="_blank" rel="noopener noreferrer" class="">支持配置 API Server 边车容器</a></p>
</li>
<li class="">
<p>支持配置组件优先级类：用户可通过自定义 Karmada CR 为控制平面组件指定优先级类，确保关键组件以适当的优先级进行调度，从而提升 Karmada 系统的可靠性和稳定性。</p>
<p>更多有关组件优先级类配置的资料请参考：<a href="https://karmada.io/docs/administrator/configuration/priority-class-configuration/#how-to-configure-priority-classes-for-karmada-control-plane-components" target="_blank" rel="noopener noreferrer" class="">如何为 Karmada 控制平面组件配置优先级类</a></p>
</li>
</ul>
<p>以上新特性进一步提升了 Karmada Operator 的可配置性，满足用户多样化需求。</p>
<h3 class="anchor anchorTargetStickyNavbar_Vzrq" id="karmada-控制器性能优化提升">Karmada 控制器性能优化提升<a href="https://karmada.io/zh/blog/2025/03/19/karmada-v1.13/karmada-v1.13#karmada-%E6%8E%A7%E5%88%B6%E5%99%A8%E6%80%A7%E8%83%BD%E4%BC%98%E5%8C%96%E6%8F%90%E5%8D%87" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h3>
<p>随着 Karmada 被业界广泛采纳，以及部署的规模不断提升，性能优化也成了 Karmada 关注和发力的重要方向。社区成立了性能优化团队，专门分析和优化 Karmada 关键组件的性能，并已取得了显著的成效。</p>
<p>在版本 v1.13.0，Karmada 的性能优化主要集中在控制器，减少大规模数据部署场景下，控制器重启时的性能开销。为了验证性能优化的成效，社区准备了 12C CPU, 28G RAM 的物理机，并部署了 1000 个 Deployments 和 1000个 Configmaps，共生成 2000 个 ResourceBindings。在重启组件 karmada-controller-manager 后，收集 workqueue 的指标，结果如下：</p>
<ul>
<li class=""><strong>对于 Detector 控制器：</strong> 队列 Reconcile 时间降低了60%。</li>
<li class=""><strong>对于 binding-controller 控制器：</strong> 队列 Reconcile 时间降低了25%。</li>
</ul>
<p>这些指标显示了在版本 v1.13.0，控制器 Detector 和 binding-controller 在重启场景下存量资源重新编排的性能有了显著的提升。</p>
<h3 class="anchor anchorTargetStickyNavbar_Vzrq" id="karmada-dashboard-首个版本发布">Karmada Dashboard 首个版本发布<a href="https://karmada.io/zh/blog/2025/03/19/karmada-v1.13/karmada-v1.13#karmada-dashboard-%E9%A6%96%E4%B8%AA%E7%89%88%E6%9C%AC%E5%8F%91%E5%B8%83" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h3>
<p>经过数位热心开发者的持续努力，Karmada Dashboard 终于迎来了具有里程碑意义的第一个版本（v0.1.0）！这一版本标志着 Karmada 在可视化管理领域迈出了重要的一步。</p>
<p>Karmada Dashboard 是一款专为 Karmada 用户设计的图形化界面工具，旨在简化多集群管理的操作流程，提升用户体验。通过 Dashboard，用户可以直观地查看集群状态、资源分布以及任务执行情况，同时还能轻松完成配置调整和策略部署。</p>
<p>Karmada Dashboard v0.1.0 主要功能包括：</p>
<ul>
<li class="">集群管理： 提供集群接入和状态概览，包括健康状态、节点数量等关键指标。</li>
<li class="">资源管理：业务资源配置管理，包括命名空间、工作负载、服务等。</li>
<li class="">策略管理：Karmada 策略管理，包括分发策略、差异化策略等。</li>
</ul>
<p>更多有关 Karmada Dashboard 的资料请参考：<a href="https://github.com/karmada-io/dashboard" target="_blank" rel="noopener noreferrer" class="">Karmada Dashboard</a></p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="致谢贡献者">致谢贡献者<a href="https://karmada.io/zh/blog/2025/03/19/karmada-v1.13/karmada-v1.13#%E8%87%B4%E8%B0%A2%E8%B4%A1%E7%8C%AE%E8%80%85" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p>Karmada v1.13 版本包含了来自 41 位贡献者的 205 次代码提交，在此对各位贡献者表示由衷的感谢：</p>
<table><thead><tr><th>^-^</th><th>^-^</th><th>^-^</th></tr></thead><tbody><tr><td>@adwait-godbole</td><td>@anujagrawal699</td><td>@axif0</td></tr><tr><td>@carlory</td><td>@chaosi-zju</td><td>@CharlesQQ</td></tr><tr><td>@chouchongYHMing</td><td>@devadapter</td><td>@dongjiang1989</td></tr><tr><td>@gabrielsrs</td><td>@guozheng-shen</td><td>@Heylosky</td></tr><tr><td>@iawia002</td><td>@jabellard</td><td>@jhnine</td></tr><tr><td>@JimDevil</td><td>@LavredisG</td><td>@LeonZh0u</td></tr><tr><td>@LiZhenCheng9527</td><td>@ls-2018</td><td>@mohamedawnallah</td></tr><tr><td>@Monokaix</td><td>@mszacillo</td><td>@RainbowMango</td></tr><tr><td>@sachinparihar</td><td>@samzong</td><td>@seanlaii</td></tr><tr><td>@shauvet</td><td>@SkySingh04</td><td>@tiansuo114</td></tr><tr><td>@Vacant2333</td><td>@vibgreon</td><td>@warjiang</td></tr><tr><td>@whitewindmills</td><td>@XiShanYongYe-Chang</td><td>@y1hao</td></tr><tr><td>@yashpandey06</td><td>@zach593</td><td>@zhouqunjie-cs</td></tr><tr><td>@zhzhuang-zju</td><td>@ZwangaMukwevho</td><td></td></tr></tbody></table>
<p><img decoding="async" loading="lazy" alt="karmada v1.13 contributors" src="https://karmada.io/zh/assets/images/contributors-64025fcd8476635d8fb07153cf8fa351.png" width="532" height="640" class="img_ev3q"></p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Karmada 安全审计结果发布]]></title>
            <link>https://karmada.io/zh/blog/2025/01/17/karmada-security-audit/karmada-security-audit</link>
            <guid>https://karmada.io/zh/blog/2025/01/17/karmada-security-audit/karmada-security-audit</guid>
            <pubDate>Fri, 17 Jan 2025 00:00:00 GMT</pubDate>
            <description><![CDATA[社区文章同步发布于 OSTIF 博客和 CNCF 博客。]]></description>
            <content:encoded><![CDATA[<p>社区文章同步发布于 <a href="https://ostif.org/karmada-audit-complete/" target="_blank" rel="noopener noreferrer" class="">OSTIF 博客</a>和 <a href="https://www.cncf.io/blog/2025/01/16/announcing-the-results-of-the-karmada-security-audit/" target="_blank" rel="noopener noreferrer" class="">CNCF 博客</a>。</p>
<p><a href="https://ostif.org/" target="_blank" rel="noopener noreferrer" class="">OSTIF</a> 非常荣幸能够分享我们对 Karmada 的安全审计结果。Karmada 是一个开源的 Kubernetes 编排系统，用于跨云和集群无缝运行云原生应用程序。
在 <a href="https://www.shielder.com/" target="_blank" rel="noopener noreferrer" class="">Shielder</a> 和 <a href="https://www.cncf.io/" target="_blank" rel="noopener noreferrer" class="">云原生计算基金会 (CNCF)</a> 的帮助下，此项目为用户提供了一个更加安全可靠的开放的多云、多集群 Kubernetes 管理解决方案。</p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="审计流程">审计流程：<a href="https://karmada.io/zh/blog/2025/01/17/karmada-security-audit/karmada-security-audit#%E5%AE%A1%E8%AE%A1%E6%B5%81%E7%A8%8B" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p>Karmada 是 Kubernetes 生态系统的一部分，使用了 Kubernetes 库和实现，除此之外，Karmada 自定义实现及其第三方依赖项的整体安全状况也是本次审计工作的重中之重。
Karmada 利用多个组件、CLI 工具和附加组件来扩展标准 Kubernetes 功能，这些功能可以根据部署配置进行定制。因此，Karmada 的攻击场景相对复杂，有必要执行范围威胁建模以评估潜在的攻击面。利用这个定制的威胁模型，结合手动检查、工具分析和动态审查，<a href="https://www.shielder.com/" target="_blank" rel="noopener noreferrer" class="">Shielder</a>识别了六个对项目安全有影响的问题。</p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="审计结果">审计结果：<a href="https://karmada.io/zh/blog/2025/01/17/karmada-security-audit/karmada-security-audit#%E5%AE%A1%E8%AE%A1%E7%BB%93%E6%9E%9C" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<ul>
<li class="">6 个发现<!-- -->
<ul>
<li class="">1 个高风险，1 个中风险，2 个低风险，2 个提示</li>
</ul>
</li>
<li class="">对未来工作的建议</li>
<li class="">整体安全性的长期改善建议</li>
</ul>
<p>Karmada 项目的安全团队在整个审计过程中一直积极响应并与 Shielder 积极合作，解决修复了报告中列出的问题。他们为项目所做的工作一丝不苟，在问题修复过程中能考虑到对用户
以及相关的第三方依赖项和项目的影响。他们发布了必要的安全通告，并告知用户本次审计的影响和提供相应的解决方案。OSTIF 祝他们在 CNCF 毕业之路上一切顺利。</p>
<p><strong>感谢以下个人和团体使这次合作成为可能：</strong></p>
<ul>
<li class="">Karmada 维护者和社区：特别是 Kevin Wang、Hongcai Ren 和 Zhuang Zhang</li>
<li class="">Shielder: Abdel Adim “Smaury” Oisfi, Pietro Tirenna, Davide Silvetti</li>
<li class="">云原生计算基金会</li>
</ul>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="参考资料">参考资料：<a href="https://karmada.io/zh/blog/2025/01/17/karmada-security-audit/karmada-security-audit#%E5%8F%82%E8%80%83%E8%B5%84%E6%96%99" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<ol>
<li class="">CNCF (宣布 Karmada 安全审计结果): <a href="https://www.cncf.io/blog/2025/01/16/announcing-the-results-of-the-karmada-security-audit/" target="_blank" rel="noopener noreferrer" class="">https://www.cncf.io/blog/2025/01/16/announcing-the-results-of-the-karmada-security-audit/</a></li>
<li class="">审计报告: <a href="https://ostif.org/wp-content/uploads/2025/01/OSTIF-Karmada-Report-PT-v1.1.pdf" target="_blank" rel="noopener noreferrer" class="">https://ostif.org/wp-content/uploads/2025/01/OSTIF-Karmada-Report-PT-v1.1.pdf</a></li>
<li class="">Shielder: <a href="https://www.shielder.com/blog/2025/01/karmada-security-audit/" target="_blank" rel="noopener noreferrer" class="">https://www.shielder.com/blog/2025/01/karmada-security-audit/</a></li>
</ol>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Karmada v1.12 版本发布！单集群应用迁移可维护性增强]]></title>
            <link>https://karmada.io/zh/blog/2024/12/23/karmada-v1.12/karmada-v1.12</link>
            <guid>https://karmada.io/zh/blog/2024/12/23/karmada-v1.12/karmada-v1.12</guid>
            <pubDate>Mon, 23 Dec 2024 00:00:00 GMT</pubDate>
            <description><![CDATA[Karmada 是开放的多云多集群容器编排引擎，旨在帮助用户在多云环境下部署和运维业务应用。凭借兼容 Kubernetes 原生 API 的能力，Karmada 可以平滑迁移单集群工作负载，并且仍可保持与 Kubernetes 周边生态工具链协同。]]></description>
            <content:encoded><![CDATA[<p>Karmada 是开放的多云多集群容器编排引擎，旨在帮助用户在多云环境下部署和运维业务应用。凭借兼容 Kubernetes 原生 API 的能力，Karmada 可以平滑迁移单集群工作负载，并且仍可保持与 Kubernetes 周边生态工具链协同。</p>
<p><img decoding="async" loading="lazy" alt="karmada" src="https://karmada.io/zh/assets/images/karmada-8385b5994e0116918f15450ccd9ba027.png" width="524" height="174" class="img_ev3q"></p>
<p><a href="https://github.com/karmada-io/karmada/releases/tag/v1.12.0" target="_blank" rel="noopener noreferrer" class="">Karmada v1.12版本</a> 现已发布，本版本包含下列新增特性：</p>
<ul>
<li class="">应用级故障迁移功能增强（新增状态中继机制，适用于大数据处理程序高可用场景，如 Flink）</li>
<li class="">单集群应用迁移能力增强（适用于单集群存量应用迁移）</li>
<li class="">Karmada Operator 高可用部署能力支持</li>
<li class="">OverridePolicy 支持局部修改结构化字段值</li>
</ul>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="新特性概览">新特性概览<a href="https://karmada.io/zh/blog/2024/12/23/karmada-v1.12/karmada-v1.12#%E6%96%B0%E7%89%B9%E6%80%A7%E6%A6%82%E8%A7%88" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<h3 class="anchor anchorTargetStickyNavbar_Vzrq" id="应用级故障迁移功��能增强">应用级故障迁移功能增强<a href="https://karmada.io/zh/blog/2024/12/23/karmada-v1.12/karmada-v1.12#%E5%BA%94%E7%94%A8%E7%BA%A7%E6%95%85%E9%9A%9C%E8%BF%81%E7%A7%BB%E5%8A%9F%E8%83%BD%E5%A2%9E%E5%BC%BA" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h3>
<p>在之前的版本中，Karmada 提供了基本的应用级故障迁移能力，能够通过应用的健康状态或自定义的故障等条件触发应用迁移。为了满足有状态应用在故障迁移过程中保留其运行状态的需求，Karmada 在 v1.12 版本新增了应用状态中继机制。对于大数据处理应用（例如 Flink），利用此能力可以从故障前的 checkpoint 重新启动，无缝恢复到重启前的数据处理状态，从而避免数据重复处理。</p>
<p>社区在PropagationPolicy/ClusterPropagationPolicy API 中的.spec.failover.application 下引入了一个新的StatePreservation 字段， 用于定义有状态应用在故障迁移期间保留和恢复状态数据的策略。结合此策略，当应用从一个故障集群迁移到另一个集群时，能够从原始资源配置中提取关键数据。</p>
<p>状态保留策略StatePreservation 包含了一系列StatePreservationRule 配置，通过 JSONPath 来指定需要保留的状态数据片段，并利用关联的AliasLabelName 将数据传递到迁移后的集群。</p>
<p>以 Flink 应用为例，在 Flink 应用中，jobID 是一个唯一的标识符，用于区分和管理不同的 Flink 作业（jobs）。每个 Flink 作业在提交到 Flink 集群时都会被分配一个jobID。当作业发生故障时，Flink 应用可以利用jobID 来恢复故障前作业的状态，从故障点处继续执行。具体的配置和步骤如下：</p>
<div class="language-yaml codeBlockContainer_Ckt0 theme-code-block" style="--prism-color:#393A34;--prism-background-color:#f6f8fa"><div class="codeBlockContent_QJqH"><pre tabindex="0" class="prism-code language-yaml codeBlock_bY9V thin-scrollbar" style="color:#393A34;background-color:#f6f8fa"><code class="codeBlockLines_e6Vv"><span class="token-line" style="color:#393A34"><span class="token key atrule" style="color:#00a4db">apiVersion</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> policy.karmada.io/v1alpha1</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">kind</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> PropagationPolicy</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">metadata</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">name</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> foo</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">spec</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token comment" style="color:#999988;font-style:italic"># ...</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">failover</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">application</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">decisionConditions</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">        </span><span class="token key atrule" style="color:#00a4db">tolerationSeconds</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token number" style="color:#36acaa">60</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">purgeMode</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> Immediately</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">statePreservation</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">        </span><span class="token key atrule" style="color:#00a4db">rules</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">          </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">aliasLabelName</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> application.karmada.io/failover</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">jobid</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">            </span><span class="token key atrule" style="color:#00a4db">jsonPath</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token string" style="color:#e3116c">"{ .jobStatus.jobID }"</span><br></span></code></pre></div></div>
<p>迁移前，Karmada 控制器将按照用户配置的路径提取 job ID。
迁移时，Karmada 控制器将提取的 job ID 以 label 的形式注入到 Flink 应用配置中，比如<code>application.karmada.io/failover-jobid : &lt;jobID&gt;</code>。
运行在成员集群的 Kyverno 拦截 Flink 应用创建请求，并根据jobID  获取该 job 的 checkpoint 数据存储路径，比如  <code>/&lt;shared-path&gt;/&lt;job-namespace&gt;/&lt;jobId&gt;/checkpoints/xxx</code>，然后配置initialSavepointPath 指示从save point 启动。
Flink 应用根据initialSavepointPath 下的 checkpoint 数据启动，从而继承迁移前保存的最终状态。
该能力基于 FlinkDeployment 打造，但广泛适用于能够基于某个 save point 启动的有状态应用程序，这些应用均可参考上述流程实现故障迁移的状态中继。</p>
<p>此功能需要启用 StatefulFailoverInjection 特性开关。StatefulFailoverInjection 目前处于 Alpha 阶段，默认情况下是关闭的。</p>
<p>功能约束：</p>
<p>应用必须限定在单个集群中运行；</p>
<p>迁移清理策略（PurgeMode）限定为Immediately，即故障应用需立即删除然后再创建新应用，确保数据一致性。</p>
<h3 class="anchor anchorTargetStickyNavbar_Vzrq" id="单集群应用迁移能力增强">单集群应用迁移能力增强<a href="https://karmada.io/zh/blog/2024/12/23/karmada-v1.12/karmada-v1.12#%E5%8D%95%E9%9B%86%E7%BE%A4%E5%BA%94%E7%94%A8%E8%BF%81%E7%A7%BB%E8%83%BD%E5%8A%9B%E5%A2%9E%E5%BC%BA" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h3>
<p>在用户将业务从单集群迁移至多集群的过程中，如果资源已经被迁移到 Karmada 控制面，那么当控制面中的资源模板被删除时，成员集群中的资源也会随之删除。但在某些场景，用户希望能够保留成员集群中的资源。例如，作为管理员，在工作负载迁移过程中可能遇到意外情况（如云平台无法发布应用程序或 Pod 异常）， 需要回滚机制立刻恢复到迁移之前的状态，以便快速止损。</p>
<p>在 v1.12 版本，社区在PropagationPolicy/ClusterPropagationPolicy API 中引入了PreserveResourcesOnDeletion 字段，用于定义当控制面中的资源模板被删除时成员集群上资源的保留行为，如果设置为true，则成员集群上的资源将被保留。结合此字段，一旦用户在迁移过程中发现异常，可以快速执行回滚操作并保留成员集群中原有的资源，整个迁移回滚过程更加安全可控。</p>
<p>使用该字段请注意以下两点：</p>
<p>该配置对所有成员集群统一生效，不会仅针对部分集群进行选择性控制。
当 Policy 被删除时，资源模板及已分发的成员集群资源将保持不变，除非被显式删除。
以 PropagationPolicy 为例，用户在删除 Karmada 控制面资源模板时，可以配置如下 PropagationPolicy 来保留成员集群的资源：</p>
<div class="language-yaml codeBlockContainer_Ckt0 theme-code-block" style="--prism-color:#393A34;--prism-background-color:#f6f8fa"><div class="codeBlockContent_QJqH"><pre tabindex="0" class="prism-code language-yaml codeBlock_bY9V thin-scrollbar" style="color:#393A34;background-color:#f6f8fa"><code class="codeBlockLines_e6Vv"><span class="token-line" style="color:#393A34"><span class="token key atrule" style="color:#00a4db">apiVersion</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> policy.karmada.io/v1alpha1</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">kind</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> PropagationPolicy</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">metadata</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">name</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> nginx</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">pp</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">spec</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">conflictResolution</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> Overwrite</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">preserveResourcesOnDeletion</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token boolean important" style="color:#36acaa">true</span><span class="token plain"> </span><span class="token comment" style="color:#999988;font-style:italic"># 资源模板删除后，成员集群资源依然保留</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">placement</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">clusterAffinity</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">clusterNames</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">        </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> member1</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">resourceSelectors</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">apiVersion</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> apps/v1</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">kind</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> Deployment</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">name</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> nginx</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">apiVersion</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> v1</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">kind</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> Service</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">name</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> nginx</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">svc</span><br></span></code></pre></div></div>
<p>更多有关安全回滚迁移的资料请参考：<a href="https://karmada.io/zh/docs/next/administrator/migration/migrate-in-batch" target="_blank" rel="noopener noreferrer" class="">迁移操作如何回滚</a></p>
<h3 class="anchor anchorTargetStickyNavbar_Vzrq" id="karmada-operator-高可用部署能力支持">Karmada Operator 高可用部署能力支持<a href="https://karmada.io/zh/blog/2024/12/23/karmada-v1.12/karmada-v1.12#karmada-operator-%E9%AB%98%E5%8F%AF%E7%94%A8%E9%83%A8%E7%BD%B2%E8%83%BD%E5%8A%9B%E6%94%AF%E6%8C%81" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h3>
<p>作为社区维护的安装工具，Karmada-operator 可以用来部署和管理多个 Karmada 实例。为了更好地支持高可用部署方案，karmada-operator 在本版本实施了一系列针对性的改进和优化措施，包括：</p>
<p>引入了对自定义 CA 证书的支持；
支持连接外部 etcd；
可通过 Secret 指定外部 etcd 客户端的凭据；
可为 Karmada 组件指定卷和卷挂载；
对外暴露 APISever 服务，用于服务发现。
这些增强使得 karmada-operator 能够跨多个管理集群部署一个高度可用的 Karmada 控制平面，这些集群可以跨越不同的数据中心，从而满足故障恢复的诉求。</p>
<p><img decoding="async" loading="lazy" alt="architecture" src="https://karmada.io/zh/assets/images/architecture-26ebc11a5ce9995c84d56054cd31ae75.png" width="1080" height="944" class="img_ev3q"></p>
<p>上图是通过 Karmada-operator 构建的生产级高可用架构，在这个架构中，Karmada-operator 跨不同地理位置的数据中心部署多个 Karmada 控制面，并将它们连接到同一个外部 etcd 集群。这种设置不仅确保了跨数据中心的数据一致性，还简化了数据管理和维护工作。</p>
<p>此外，借助 Karmada-operator 提供的 APIServer 服务暴露能力，结合 Ingress 对外提供统一的服务访问。同时，利用可配置的CA证书机制，保障了 Karmada 实例与外部服务间通信的安全性。</p>
<p>此架构显著增强了系统对单个数据中心故障的抵御能力，最大限度地减少了因数据中心故障导致的服务中断风险，保证了业务连续性和用户体验的稳定性，符合严格的灾难恢复标准。</p>
<h3 class="anchor anchorTargetStickyNavbar_Vzrq" id="overridepolicy-支持局部修改结构化字段值">OverridePolicy 支持局部修改结构化字段值<a href="https://karmada.io/zh/blog/2024/12/23/karmada-v1.12/karmada-v1.12#overridepolicy-%E6%94%AF%E6%8C%81%E5%B1%80%E9%83%A8%E4%BF%AE%E6%94%B9%E7%BB%93%E6%9E%84%E5%8C%96%E5%AD%97%E6%AE%B5%E5%80%BC" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h3>
<p>OverridePolicy 允许用户针对特定集群自定义资源的覆盖策略，确保资源可以在不同环境中灵活适配和优化。Kubernetes 资源如 Secrets 和 ConfigMaps 常常会用到结构化的字段值，如 ConfigMaps 的.data 利用 YAML 格式的结构化数据承载配置信息。在实际应用中，存在只需要修改其部分字段的情况，而且，当原始的结构化字段值复杂且内容繁多时，使用全覆盖将会大大增大 OverridePolicy 的配置难度。</p>
<p>为了解决这一问题，并提高 OverridePolicy 在此类场景中的易用性，Karmada 引入了FieldOverrider 特性。FieldOverrider 支持对 JSON 和 YAML 格式的结构化字段值进行局部修改，即只添加或替换或删除所需的字段。这种方式简化了配置过程，提高了效率，同时减少了出错的可能性，使得资源管理更加直观和便捷。通过FieldOverrider，用户可以对结构化字段值进行更精细化地处理，适应多变的应用环境需求。</p>
<p>下面以 ConfigMap 为例，用户可通过FieldOverrider 部分覆盖 ConfigMap 的.data 字段来实现集群间的差异化配置。</p>
<div class="language-yaml codeBlockContainer_Ckt0 theme-code-block" style="--prism-color:#393A34;--prism-background-color:#f6f8fa"><div class="codeBlockContent_QJqH"><pre tabindex="0" class="prism-code language-yaml codeBlock_bY9V thin-scrollbar" style="color:#393A34;background-color:#f6f8fa"><code class="codeBlockLines_e6Vv"><span class="token-line" style="color:#393A34"><span class="token comment" style="color:#999988;font-style:italic"># example-configmap</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">apiVersion</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> v1</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">kind</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> ConfigMap</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">metadata</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">name</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> example</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">configmap</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">data</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">config.yaml</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token punctuation" style="color:#393A34">|</span><span class="token scalar string" style="color:#e3116c"></span><br></span><span class="token-line" style="color:#393A34"><span class="token scalar string" style="color:#e3116c">    app:</span><br></span><span class="token-line" style="color:#393A34"><span class="token scalar string" style="color:#e3116c">      database:</span><br></span><span class="token-line" style="color:#393A34"><span class="token scalar string" style="color:#e3116c">        port: 5432</span><br></span><span class="token-line" style="color:#393A34"><span class="token scalar string" style="color:#e3116c">        ip: 127.0.0.1</span><br></span><span class="token-line" style="color:#393A34"><span class="token scalar string" style="color:#e3116c">        name: example</span><br></span><span class="token-line" style="color:#393A34"><span class="token scalar string" style="color:#e3116c">        zone: zone1</span><br></span></code></pre></div></div>
<div class="language-yaml codeBlockContainer_Ckt0 theme-code-block" style="--prism-color:#393A34;--prism-background-color:#f6f8fa"><div class="codeBlockContent_QJqH"><pre tabindex="0" class="prism-code language-yaml codeBlock_bY9V thin-scrollbar" style="color:#393A34;background-color:#f6f8fa"><code class="codeBlockLines_e6Vv"><span class="token-line" style="color:#393A34"><span class="token comment" style="color:#999988;font-style:italic"># example-overridepolicy</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">apiVersion</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> policy.karmada.io/v1alpha1</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">kind</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> OverridePolicy</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">metadata</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">name</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> example</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">spec</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">resourceSelectors</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">apiVersion</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> v1</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">kind</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> ConfigMap</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">name</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> example</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">configmap</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">overrideRules</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">overriders</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">        </span><span class="token key atrule" style="color:#00a4db">fieldOverrider</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">          </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">fieldPath</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> /data/config.yaml</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">            </span><span class="token key atrule" style="color:#00a4db">yaml</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">              </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">subPath</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> /app/database/port</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">                </span><span class="token key atrule" style="color:#00a4db">operator</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> replace </span><span class="token comment" style="color:#999988;font-style:italic"># 支持add、remove和replace操作</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">                </span><span class="token key atrule" style="color:#00a4db">value</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token string" style="color:#e3116c">"3306"</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">targetCluster</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">        </span><span class="token key atrule" style="color:#00a4db">clusterNames</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">          </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> member1</span><br></span></code></pre></div></div>
<p>经过以上配置，集群 member1 中的 ConfigMap 将更新为：</p>
<div class="language-yaml codeBlockContainer_Ckt0 theme-code-block" style="--prism-color:#393A34;--prism-background-color:#f6f8fa"><div class="codeBlockContent_QJqH"><pre tabindex="0" class="prism-code language-yaml codeBlock_bY9V thin-scrollbar" style="color:#393A34;background-color:#f6f8fa"><code class="codeBlockLines_e6Vv"><span class="token-line" style="color:#393A34"><span class="token comment" style="color:#999988;font-style:italic"># example-configmap in member1</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">apiVersion</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> v1</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">kind</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> ConfigMap</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">metadata</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">name</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> myconfigmap</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">data</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">config.yaml</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token punctuation" style="color:#393A34">|</span><span class="token scalar string" style="color:#e3116c"></span><br></span><span class="token-line" style="color:#393A34"><span class="token scalar string" style="color:#e3116c">    app:</span><br></span><span class="token-line" style="color:#393A34"><span class="token scalar string" style="color:#e3116c">      database:</span><br></span><span class="token-line" style="color:#393A34"><span class="token scalar string" style="color:#e3116c">        port: 3306 # 更新了port</span><br></span><span class="token-line" style="color:#393A34"><span class="token scalar string" style="color:#e3116c">        ip: 127.0.0.1</span><br></span><span class="token-line" style="color:#393A34"><span class="token scalar string" style="color:#e3116c">        name: example</span><br></span><span class="token-line" style="color:#393A34"><span class="token scalar string" style="color:#e3116c">        zone: zone1</span><br></span></code></pre></div></div>
<p>更多FieldOverrider 的用法请参考：<a href="https://karmada.io/docs/userguide/scheduling/override-policy/#fieldoverrider" target="_blank" rel="noopener noreferrer" class="">FieldOverrider 使用指南</a></p>
<h1>致谢贡献者</h1>
<p>Karmada v1.12 版本包含了来自 33 位贡献者的 253 次代码提交，在此对各位贡献者表示由衷的感谢：</p>
<p>贡献者列表：</p>
<table><thead><tr><th>^-^</th><th>^-^</th><th>^-^</th></tr></thead><tbody><tr><td>@a7i</td><td>@ahorine</td><td>@anujagrawal699</td></tr><tr><td>@B1f030</td><td>@chaosi-zju</td><td>@CharlesQQ</td></tr><tr><td>@chaunceyjiang</td><td>@husnialhamdani</td><td>@iawia002</td></tr><tr><td>@ipsum-0320</td><td>@jabellard</td><td>@jklaw90</td></tr><tr><td>@KhalilSantana</td><td>@LavredisG</td><td>@liangyuanpeng</td></tr><tr><td>@LivingCcj</td><td>@MAVRICK-1</td><td>@mohamedawnallah</td></tr><tr><td>@mszacillo</td><td>@RainbowMango</td><td>@SataQiu</td></tr><tr><td>@seanlaii</td><td>@sophiefeifeifeiya</td><td>@tiansuo114</td></tr><tr><td>@wangxf1987</td><td>@whitewindmills</td><td>@wulemao</td></tr><tr><td>@XiShanYongYe-Chang</td><td>@xovoxy</td><td>@yanfeng1992</td></tr><tr><td>@yelshall</td><td>@zach593</td><td>@zhzhuang-zju</td></tr></tbody></table>
<p><img decoding="async" loading="lazy" alt="karmada v1.12 contributors" src="https://karmada.io/zh/assets/images/contributors-0e9b0cacb70966a6cb763748f254a9d6.png" width="544" height="570" class="img_ev3q"></p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Karmada v1.11 版本发布！新增应用跨集群滚动升级能力！]]></title>
            <link>https://karmada.io/zh/blog/2024/09/13/karmada-v1.11/karmada-v1.11</link>
            <guid>https://karmada.io/zh/blog/2024/09/13/karmada-v1.11/karmada-v1.11</guid>
            <pubDate>Fri, 13 Sep 2024 00:00:00 GMT</pubDate>
            <description><![CDATA[Karmada 是开放的多云多集群容器编排引擎，旨在帮助用户在多云环境下部署和运维业务应用。凭借兼容 Kubernetes 原生 API 的能力，Karmada 可以平滑迁移单集群工作负载，并且仍可保持与 Kubernetes 周边生态工具链协同。]]></description>
            <content:encoded><![CDATA[<p>Karmada 是开放的多云多集群容器编排引擎，旨在帮助用户在多云环境下部署和运维业务应用。凭借兼容 Kubernetes 原生 API 的能力，Karmada 可以平滑迁移单集群工作负载，并且仍可保持与 Kubernetes 周边生态工具链协同。</p>
<p>本版本包含下列新增特性：</p>
<ul>
<li class="">支持联邦应用跨集群滚动升级，使用户版本发布流程更加灵活可控；</li>
<li class="">karmadactl 新增了多项运维能力，提供独特的多集群运维体验；</li>
<li class="">为联邦工作负载提供标准化 generation 语义，使 CD 执行一步到位；</li>
<li class="">Karmada Operator 支持自定义 CRD 下载策略，使离线部署更灵活。</li>
</ul>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="联邦应用跨集群滚动升级">联邦应用跨集群滚动升级<a href="https://karmada.io/zh/blog/2024/09/13/karmada-v1.11/karmada-v1.11#%E8%81%94%E9%82%A6%E5%BA%94%E7%94%A8%E8%B7%A8%E9%9B%86%E7%BE%A4%E6%BB%9A%E5%8A%A8%E5%8D%87%E7%BA%A7" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p>在最新发布的 v1.11 版本中，Karmada 新增了联邦应用跨集群滚动升级特性。这一特性特别适用于那些部署在多个集群上的应用，使得用户在发布应用新版本时能够采用更加灵活和可控的滚动升级策略。用户可以精细地控制升级流程，确保每个集群在升级过程中都能够平滑过渡，减少对生产环境的影响。这一特性不仅提升了用户体验，也为复杂的多集群管理提供了更多的灵活性和可靠性。</p>
<p>下面通过一个示例来演示如何对联邦应用进行滚动升级：</p>
<p>假定用户已经通过 PropagationPolicy 将 Deployment 应用分发到三个成员集群中：<code>ClusterA</code>、<code>ClusterB</code>、<code>ClusterC</code>：</p>
<div class="language-yaml codeBlockContainer_Ckt0 theme-code-block" style="--prism-color:#393A34;--prism-background-color:#f6f8fa"><div class="codeBlockContent_QJqH"><pre tabindex="0" class="prism-code language-yaml codeBlock_bY9V thin-scrollbar" style="color:#393A34;background-color:#f6f8fa"><code class="codeBlockLines_e6Vv"><span class="token-line" style="color:#393A34"><span class="token key atrule" style="color:#00a4db">apiVersion</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> policy.karmada.io/v1alpha1</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">kind</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> PropagationPolicy</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">metadata</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">name</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> nginx</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">propagation</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">spec</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">resourceSelectors</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">apiVersion</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> apps/v1</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">kind</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> Deployment</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">name</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> nginx</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">placement</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">clusterAffinity</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">clusterNames</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">        </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> ClusterA</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">        </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> ClusterB</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">        </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> ClusterC</span><br></span></code></pre></div></div>
<p><img decoding="async" loading="lazy" alt="rollout step 0" src="https://karmada.io/zh/assets/images/rollout-00-5e81a84aaaa89092eaa6550869954687.png" width="606" height="376" class="img_ev3q"></p>
<p>此时 Deployment 版本为v1，为了将 Deployment 资源版本升级至 v2，您可以依次执行下列步骤。</p>
<p>首先，通过配置 PropagationPolicy 策略，暂时停止向 <code>ClusterA</code> 和 <code>ClusterB</code> 分发资源，从而应用的变更将只发生在 <code>ClusterC</code>：</p>
<div class="language-yaml codeBlockContainer_Ckt0 theme-code-block" style="--prism-color:#393A34;--prism-background-color:#f6f8fa"><div class="codeBlockContent_QJqH"><pre tabindex="0" class="prism-code language-yaml codeBlock_bY9V thin-scrollbar" style="color:#393A34;background-color:#f6f8fa"><code class="codeBlockLines_e6Vv"><span class="token-line" style="color:#393A34"><span class="token key atrule" style="color:#00a4db">apiVersion</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> policy.karmada.io/v1alpha1</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">kind</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> PropagationPolicy</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">metadata</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">name</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> nginx</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">propagation</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">spec</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token comment" style="color:#999988;font-style:italic">#...</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">suspension</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">dispatchingOnClusters</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">clusterNames</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">        </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> ClusterA</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">        </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> ClusterB</span><br></span></code></pre></div></div>
<p><img decoding="async" loading="lazy" alt="rollout step 1" src="https://karmada.io/zh/assets/images/rollout-01-81ccba746ad00bc75fbca71ff7173093.png" width="604" height="372" class="img_ev3q"></p>
<p>然后更新 PropagationPolicy 资源，允许系统向 <code>ClusterB</code> 集群同步新版本资源：</p>
<div class="language-yaml codeBlockContainer_Ckt0 theme-code-block" style="--prism-color:#393A34;--prism-background-color:#f6f8fa"><div class="codeBlockContent_QJqH"><pre tabindex="0" class="prism-code language-yaml codeBlock_bY9V thin-scrollbar" style="color:#393A34;background-color:#f6f8fa"><code class="codeBlockLines_e6Vv"><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">suspension</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">dispatchingOnClusters</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">clusterNames</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">        </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> ClusterA</span><br></span></code></pre></div></div>
<p><img decoding="async" loading="lazy" alt="rollout step 2" src="https://karmada.io/zh/assets/images/rollout-02-d4909bfd981cf0ea2bb07c0aae4759b3.png" width="597" height="370" class="img_ev3q"></p>
<p>最后删除 PropagationPolicy 资源中的 <code>suspension</code> 字段，允许系统向 <code>ClusterA</code> 集群同步新版本资源：</p>
<p><img decoding="async" loading="lazy" alt="rollout step 3" src="https://karmada.io/zh/assets/images/rollout-03-5cb74456ca4e634719c32376ef926143.png" width="601" height="376" class="img_ev3q"></p>
<p>从上述示例中我们可以看到，利用联邦应用跨集群滚动发布能力，新版本应用可以做到按集群粒度滚动升级，并且可以做到精准控制。</p>
<p>此外，该功能还可以应用于其他场景，特别是对于开发者来说，由于 Karmada 控制平面和成员集群之间争夺资源控制权而导致资源频繁更新的情况。在这种情况下，暂停资源与成员集群的同步过程可以方便快速识别问题。</p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="karmadactl-能力增强">karmadactl 能力增强<a href="https://karmada.io/zh/blog/2024/09/13/karmada-v1.11/karmada-v1.11#karmadactl-%E8%83%BD%E5%8A%9B%E5%A2%9E%E5%BC%BA" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p>在本版本中，Karmada 社区致力于增强 Karmadactl 的能力，以便提供更好的多集群运维体验，进而摆脱用户对 kubectl 的依赖。</p>
<h3 class="anchor anchorTargetStickyNavbar_Vzrq" id="更丰富的命令集">更丰富的命令集<a href="https://karmada.io/zh/blog/2024/09/13/karmada-v1.11/karmada-v1.11#%E6%9B%B4%E4%B8%B0%E5%AF%8C%E7%9A%84%E5%91%BD%E4%BB%A4%E9%9B%86" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h3>
<p>Karmadactl 支持更丰富的命令集，如 <code>create</code>、<code>patch</code>、<code>delete</code>、<code>label</code>、<code>annotate</code>、<code>edit</code>、<code>attach</code>、<code>top node</code>、<code>api-resources</code> 以及 <code>explain</code>，这些命令允许用户对 Karmada 控制面或成员集群上的资源执行更多操作。</p>
<h3 class="anchor anchorTargetStickyNavbar_Vzrq" id="更丰富的功能">更丰富的功能<a href="https://karmada.io/zh/blog/2024/09/13/karmada-v1.11/karmada-v1.11#%E6%9B%B4%E4%B8%B0%E5%AF%8C%E7%9A%84%E5%8A%9F%E8%83%BD" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h3>
<p>Karmadactl 引入了 <code>--operation-scope</code> 参数来控制命令的操作范围。有了这个新参数，<code>get</code>、<code>describe</code>、<code>exec</code> 和 <code>explain</code> 等命令可以灵活切换集群视角对 Karmada 控制面或成员集群的资源进行操作。</p>
<h3 class="anchor anchorTargetStickyNavbar_Vzrq" id="更详细的命令输出信息">更详细的命令输出信息<a href="https://karmada.io/zh/blog/2024/09/13/karmada-v1.11/karmada-v1.11#%E6%9B%B4%E8%AF%A6%E7%BB%86%E7%9A%84%E5%91%BD%E4%BB%A4%E8%BE%93%E5%87%BA%E4%BF%A1%E6%81%AF" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h3>
<p><code>karmadactl get cluster</code> 命令的输出现在增加了 cluster 对象的 <code>Zones</code>、<code>Region</code>、<code>Provider</code>、<code>API-Endpoint</code> 和 <code>Proxy-URL</code> 信息。</p>
<p>通过这些能力增强，karmadactl 的操作和运维体验得到了提升。karmadactl 的新功能和更多详细信息可以通过使用 <code>karmadactl --help</code> 获得。</p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="联邦工作负载标准化-generation-语义">联邦工作负载标准化 generation 语义<a href="https://karmada.io/zh/blog/2024/09/13/karmada-v1.11/karmada-v1.11#%E8%81%94%E9%82%A6%E5%B7%A5%E4%BD%9C%E8%B4%9F%E8%BD%BD%E6%A0%87%E5%87%86%E5%8C%96-generation-%E8%AF%AD%E4%B9%89" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p>在本版本中，Karmada 将联邦层面的工作负载 generation 语义进行了标准化。这一更新为发布系统提供了可靠的参考，增强了跨集群部署的精确度。通过标准化 generation 语义，Karmada 简化了发布流程，并确保一致性地跟踪工作负载状态，使得跨多个集群管理和监控应用程序变得更加容易。</p>
<p>标准化细节为，当且仅当工作负载分发至所有成员集群中的资源状态满足 <code>status.observedGeneration</code> &gt;= <code>metadata.generation</code> 时，联邦层面的工作负载状态中的 observedGeneration 值才会被设置为其本身 <code>.metadata.generation</code> 值，这确保了每个成员集群中相应的控制器均已完成了对该工作负载的处理。此举将联邦层面的 generation 语义同kubernetes 集群的 generation 语义进行了统一，使用户能够更便捷的将单集群业务迁移至多集群业务。</p>
<p>本版本已完成下列资源适配：</p>
<ul>
<li class="">GroupVersion: apps/v1 Kind: Deployment, DaemonSet, StatefulSet</li>
<li class="">GroupVersion: apps.kruise.io/v1alpha1 Kind: CloneSet, DaemonSet</li>
<li class="">GroupVersion: apps.kruise.io/v1beta1 Kind: StatefulSet</li>
<li class="">GroupVersion: helm.toolkit.fluxcd.io/v2beta1 Kind: HelmRelease</li>
<li class="">GroupVersion: kustomize.toolkit.fluxcd.io/v1 Kind: Kustomization</li>
<li class="">GroupVersion: source.toolkit.fluxcd.io/v1 Kind: GitRepository</li>
<li class="">GroupVersion: source.toolkit.fluxcd.io/v1beta2 Kind: Bucket, HelmChart, HelmRepository, OCIRepository</li>
</ul>
<p>如有您有更多资源（包括CRD）需要适配，可以向 Karmada 社区进行反馈，也可以使用 <a href="https://karmada.io/docs/next/userguide/globalview/customizing-resource-interpreter/" target="_blank" rel="noopener noreferrer" class="">Resource Interpreter</a> 进行扩展。</p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="karmada-operator-支持自定义-crd-下载策略">Karmada Operator 支持自定义 CRD 下载策略<a href="https://karmada.io/zh/blog/2024/09/13/karmada-v1.11/karmada-v1.11#karmada-operator-%E6%94%AF%E6%8C%81%E8%87%AA%E5%AE%9A%E4%B9%89-crd-%E4%B8%8B%E8%BD%BD%E7%AD%96%E7%95%A5" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p>CRD（Custom Resource Definition，自定义资源定义）资源是 Karmada Operator 用于配置新的 Karmada 实例的关键前提资源。这些 CRD 资源包含了 Karmada 系统的关键 API 定义，例如，PropagationPolicy，ResourceBinding，Work 等。</p>
<p>在 v.1.11 版本中，Karmada Operator 支持用户自定义 CRD 下载策略。利用这个功能，用户可以指定 CRD 资源的下载路径，并定义更多的下载策略，为用户提供了更灵活的离线部署方式。</p>
<p>有关该特性的详细描述，可以参考<a href="https://github.com/karmada-io/karmada/tree/master/docs/proposals/operator-custom-crd-download-strategy" target="_blank" rel="noopener noreferrer" class="">提案：Custom CRD Download Strategy Support for Karmada Operator</a>。</p>
<h1>致谢贡献者</h1>
<p>Karmada v1.11 版本包含了来自 36 位贡献者的 223 次代码提交，在此对各位贡献者表示由衷的感谢：</p>
<p><img decoding="async" loading="lazy" alt="karmada v1.11 contributors" src="https://karmada.io/zh/assets/images/contributors-4c5f406e212580e713f76e71cba9ffdd.png" width="968" height="1092" class="img_ev3q"></p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[突破100倍集群规模，Karmada大规模测试报告发布]]></title>
            <link>https://karmada.io/zh/blog/2022/10/26/test-report</link>
            <guid>https://karmada.io/zh/blog/2022/10/26/test-report</guid>
            <pubDate>Wed, 26 Oct 2022 00:00:00 GMT</pubDate>
            <description><![CDATA[摘要]]></description>
            <content:encoded><![CDATA[<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="摘要">摘要<a href="https://karmada.io/zh/blog/2022/10/26/test-report#%E6%91%98%E8%A6%81" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p>随着云原生技术在越来越多的企业和组织中的大规模落地，如何高效、可靠地管理大规模资源池以应对不断增长的业务挑战成为了当下云原生技术的关键挑战。
在过去的很长一段时间内，不同厂商尝试通过定制Kubernetes原生组件的方式扩展单集群的规模，这在扩大规模的同时也引入了复杂的单集群运维、不清晰的集群升级路径等问题。
而多集群技术能在不侵入修改Kubernetes单集群的基础上横向扩展资源池的规模，在扩展资源池的同时降低了企业的运维管理等成本。</p>
<p>在Karmada的大规模落地进程中，Karmada的可扩展性和大规模逐渐成为社区用户的新关注点。
因此，我们对Karmada开展了大规模环境下的测试工作，以获取Karmada管理多个Kubernetes集群的性能基线指标。
<strong>对于以Karmada为代表的多集群系统而言，单集群的规模不是制约它的资源池规模的限制因素。</strong>
因此，我们参考了<strong>Kubernetes的大规模集群的标准配置和用户的生产落地实践</strong>，测试了Karmada同时管理100个5k节点和2wPod的Kubernetes集群的用户场景。
受限于测试环境和测试工具，本次测试并未追求测试到Karmada多集群系统的上限，而是希望能覆盖到在生产中大规模使用多集群技术的典型场景。
根据测试结果分析，以Karmada为核心的集群联邦可以<strong>稳定支持100个大规模集群</strong>，管理超过50万个节点和200万个Pod，可以满足用户在大规模生产落地的需要。</p>
<p>在本文中，我们将介绍用于测试的相关指标，如何进行大规模测试，以及我们如何实现大规模的集群接入。</p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="背景">背景<a href="https://karmada.io/zh/blog/2022/10/26/test-report#%E8%83%8C%E6%99%AF" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p>随着云原生技术的不断发展和使用场景的不断丰富，多云、分布式云逐渐成为引领云计算发展的趋势。著名分析公司Flexera在2021的调查报告显示，超过93%的企业正同时使用多个云厂商的服务，一方面受限于Kubernetes单集群的业务承载能力和故障恢复能力，单一的集群无法适应现有的企业业务，另一方面，在全球化的当下，企业出于避免被单家厂商垄断的目的，或是出于成本等因素考虑，更倾向于选择混合云或者多公有云的架构。 与此同时，Karmada社区的用户在落地的进程中也提出了多集群下大规模节点和应用管理的诉求。</p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="karmada介绍">Karmada介绍<a href="https://karmada.io/zh/blog/2022/10/26/test-report#karmada%E4%BB%8B%E7%BB%8D" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p>Karmada(Kubernetes Armada)是一个Kubernetes管理系统，它能够使你在无需修改应用的情况下跨集群和跨云运行你的云原生应用。通过使用Kubernetes原生API并在其上提供高级调度功能，Karmada实现了真正开放的多云Kubernetes。</p>
<p>Karmada旨在为多云和混合云场景中的多集群应用管理提供完全的自动化。它具备集中式多云管理、高可用性、故障恢复和流量调度等关键特性。</p>
<p><img decoding="async" loading="lazy" alt="architecture" src="https://karmada.io/zh/assets/images/architecture-37447d3b4fceeae700e488373138d808.png" width="551" height="487" class="img_ev3q"></p>
<p>Karmada控制面包括以下组件:</p>
<ul>
<li class="">Karmada API Server</li>
<li class="">Karmada Controller Manager</li>
<li class="">Karmada Scheduler</li>
</ul>
<p>ETCD存储了Karmada的API对象, karmada-apiserver提供了与所有其他组件通信的REST端口, 之后由karmada-controller-manager对你向karmada-apiserver提交的API对象进行对应的调和操作。</p>
<p>karmada-controller-manager运行着各种控制器，这些控制器watch着Karmada的对象，然后发送请求至成员集群的apiserver来创建常规的Kubernetes资源。</p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="多集群系统资源池规模的维度和阈值">多集群系统资源池规模的维度和阈值<a href="https://karmada.io/zh/blog/2022/10/26/test-report#%E5%A4%9A%E9%9B%86%E7%BE%A4%E7%B3%BB%E7%BB%9F%E8%B5%84%E6%BA%90%E6%B1%A0%E8%A7%84%E6%A8%A1%E7%9A%84%E7%BB%B4%E5%BA%A6%E5%92%8C%E9%98%88%E5%80%BC" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p>一个多集群系统的资源池规模不单指集群数量，即Scalability!=#Num of Clusters, 实际上多集群资源池规模包含很多维度的测量，在不考虑其他维度的情况下只考虑集群数量是毫无意义的。</p>
<p>我们将一个多集群的资源池规模<strong>按优先级</strong>描述为以下所示的三个维度：</p>
<ol>
<li class="">Num of Clusters: 集群数量是衡量一个多集群系统资源池规模和承载能力最直接且最重要的维度，在其余维度不变的情况下系统能接入的集群数量越多，说明系统的资源池规模越大，承载能力越强。</li>
<li class="">Num of Resources(API Objects): 对于一个多集群系统的控制面来说，存储并不是无限制的，而在控制面创建的资源对象的数量和总体大小受限于系统控制面的存储，也是制约多集群系统资源池规模的重要维度。这里的资源对象不仅指下发到成员集群的资源模板，而且还包括集群的调度策略、多集群服务等资源。</li>
<li class="">Cluster Size: 集群规模是衡量一个多集群系统资源池规模不可忽视的维度。一方面，集群数量相等的情况下，单个集群的规模越大，整个多集群系统的资源池越大。另一方面，多集群系统的上层能力依赖系统对集群的资源画像，例如在多集群应用的调度过程中，集群资源是不可或缺的一个因素。 <strong>综上所述，单集群的规模与整个多集群系统息息相关，但单集群的规模同样不是制约多集群系统的限制因素</strong>。用户可以通过优化原生的Kubernetes组件的方式来提升单集群的集群规模，达到扩大整个多集群系统的资源池的目的，但这不是衡量多集群系统性能的关注点。
本次测试中，社区参考了kubernetes的<a href="https://kubernetes.io/zh-cn/docs/setup/best-practices/cluster-large/" target="_blank" rel="noopener noreferrer" class="">大规模集群的标准配置</a>以及测试工具的性能，制定了测试集群的规模，以贴切实际生产环境中的单集群配置。在集群的标准配置中，Node与Pod毫无疑问是其中最重要的两个资源，Node是计算、存储等资源的最小载体，而Pod数量则代表着一个集群的应用承载能力。
事实上，单集群的资源对象也包括像service，configmap，secret这样的常见资源。这些变量的引入会使得测试过程变得更复杂，所以这次测试不会过多关注这些变量。<!-- -->
<ul>
<li class="">Num of Nodes</li>
<li class="">Num of Pods</li>
</ul>
</li>
</ol>
<p>对于多集群系统而言想要无限制地扩展各个维度而且又满足SLIs/SLOs各项指标显然是不可能实现的。
各个维度不是完全独立的，某个维度被拉伸相应的其他维度就要被压缩，可以根据使用场景进行调整。
以Clusters和Nodes两个维度举例，在100集群下将单集群的5k 节点拉伸到10k node的场景或者在单集群规格不变的同时扩展集群数量到200集群，其他维度的规格势必会受到影响。
如果各种场景都进行测试分析工作量是非常巨大的，在本次测试中，我们会重点选取典型场景配置进行测试分析。在满足SLIs/SLOs的基础上，实现单集群支持5k节点，20k pod规模的100数量的集群接入和管理。</p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="slisslos">SLIs/SLOs<a href="https://karmada.io/zh/blog/2022/10/26/test-report#slisslos" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p>可扩展性和性能是多集群联邦的重要特性。作为多集群联邦的用户，我们期望在以上两方面有服务质量的保证。在进行大规模性能测试之前，我们需要定义测量指标。
在参考了Kubernetes社区的SLI(Service Level Indicator)/SLO(Service Level Objectives)和多集群的典型应用，Karmada社区定义了以下SLI/SLO来衡量多集群联邦的服务质量。</p>
<ol>
<li class="">API Call Latency</li>
</ol>
<table><thead><tr><th>Status</th><th>SLI</th><th>SLO</th></tr></thead><tbody><tr><td>Official</td><td>最近5min对单个资源对象Mutating API调用（包括POST、PUT、DELETE、PATCH）的P99时延</td><td><code>P99 &lt;= 1s</code></td></tr><tr><td>Official</td><td>最近5min的non-streaming的只读API调用（包括GET和LIST）的P99时延</td><td>(a)Scope=resource, <code>P99 &lt;= 1s</code>, (b)Scope=namespace or Scope=cluster, P99 <code>&lt;=</code> 30s</td></tr></tbody></table>
<ol start="2">
<li class="">Resource Distribution Latency</li>
</ol>
<table><thead><tr><th>Status</th><th>SLI</th><th>SLO</th></tr></thead><tbody><tr><td>Official</td><td>用户在联邦控制面提交资源模板和下发策略后到资源在成员集群上被创建的P99时延，不考虑控制面与成员集群之间的网络波动</td><td><code>P99 &lt;= 2s</code></td></tr></tbody></table>
<ol start="3">
<li class="">Cluster Registration Latency</li>
</ol>
<table><thead><tr><th>Status</th><th>SLI</th><th>SLO</th></tr></thead><tbody><tr><td>WIP</td><td>集群从接入联邦控制面到状态能被控制面正常收集的P99时延，不考虑控制面与成员集群之间的网络波动</td><td>TBD</td></tr></tbody></table>
<ol start="4">
<li class="">Resource usage</li>
</ol>
<table><thead><tr><th>Status</th><th>SLI</th><th>SLO</th></tr></thead><tbody><tr><td>WIP</td><td>在接入一定数量的集群后集群联邦维持其正常工作所必需的资源使用量</td><td>TBD</td></tr></tbody></table>
<p><strong>Note</strong>:</p>
<ol>
<li class="">上述指标不考虑控制面和成员集群的网络波动。同时，单集群内的SLO不会考虑在内。</li>
<li class="">资源使用量是一个对于多集群系统非常重要的指标，但是不同多集群系统提供的上层服务不同，所以对各个系统来说资源的要求也会不同。我们不对这个指标进行强制的限制。</li>
<li class="">集群注册时延是从集群注册到控制面到集群在联邦侧可用的时延。它在某种程度上取决于控制面如何收集成员集群的状态。</li>
</ol>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="测试工具">测试工具<a href="https://karmada.io/zh/blog/2022/10/26/test-report#%E6%B5%8B%E8%AF%95%E5%B7%A5%E5%85%B7" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<h3 class="anchor anchorTargetStickyNavbar_Vzrq" id="clusterloader2">ClusterLoader2<a href="https://karmada.io/zh/blog/2022/10/26/test-report#clusterloader2" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h3>
<p>ClusterLoader2是一款开源Kubernetes集群负载测试工具，该工具能够针对Kubernetes 定义的SLIs/SLOs 指标进行测试，检验集群是否符合各项服务质量标准。此外ClusterLoader2为集群问题定位和集群性能优化提供可视化数据。ClusterLoader2 最终会输出一份Kubernetes集群性能报告，展示一系列性能指标测试结果。
然而，在Karmada性能测试的过程中，由于ClusterLoader2是一个为Kubernetes单集群定制的测试工具，且在多集群场景下它不能获取到所有集群的资源， 因此我们只用ClusterLoader2来分发被Karmada管理的资源。</p>
<h3 class="anchor anchorTargetStickyNavbar_Vzrq" id="prometheus">Prometheus<a href="https://karmada.io/zh/blog/2022/10/26/test-report#prometheus" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h3>
<p>Prometheus是一个开源的用于监控和告警的工具, 它包含数据收集、数据报告、数据可视化等功能。在分析了Clusterloader2对各种监控指标的处理后，我们使用Prometheus根据具体的查询语句对控制面的指标进行监控。</p>
<h3 class="anchor anchorTargetStickyNavbar_Vzrq" id="kind">Kind<a href="https://karmada.io/zh/blog/2022/10/26/test-report#kind" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h3>
<p>Kind是一个是用容器来运行Kubernetes本地集群的工具。为了测试Karmada的应用分发能力，我们需要一个真实的单集群控制面来管理被联邦控制面分发的应用。Kind能够在节约资源的同时模拟一个真实的集群。</p>
<h3 class="anchor anchorTargetStickyNavbar_Vzrq" id="fake-kubelet">Fake-kubelet<a href="https://karmada.io/zh/blog/2022/10/26/test-report#fake-kubelet" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h3>
<p>Fake-kubelet是一个能模拟节点且能维护虚拟节点上的Pod的工具。与Kubemark相比，fake-kubelet只做维护节点和Pod的必要工作。它非常适合模拟大规模的节点和Pod来测试控制面的在大规模环境下的性能。</p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="测试集群部署方案">测试集群部署方案<a href="https://karmada.io/zh/blog/2022/10/26/test-report#%E6%B5%8B%E8%AF%95%E9%9B%86%E7%BE%A4%E9%83%A8%E7%BD%B2%E6%96%B9%E6%A1%88" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p><img decoding="async" loading="lazy" alt="test architecture" src="https://karmada.io/zh/assets/images/test_architecture-3149019cdc219997d121c6654899b66d.png" width="661" height="1029" class="img_ev3q"></p>
<p>Kubernetes控制面部署在单master的节点上。etcd，kube-apiserver，kube-scheduler和kube-controller以单实例的形式部署。
Karmada的控制面组件部署在这个master节点上。他们同样以单实例的形式部署。
所有的Kubernetes组件和Karmada组件运行在高性能的节点上，且我们不对他们限制资源。
我们通过kind来模拟单master节点的集群，通过fake-kubelet来模拟集群中的工作节点。</p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="测试环境信息">测试环境信息<a href="https://karmada.io/zh/blog/2022/10/26/test-report#%E6%B5%8B%E8%AF%95%E7%8E%AF%E5%A2%83%E4%BF%A1%E6%81%AF" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p>控制面操作系统版本</p>
<p><code>Ubuntu 18.04.6 LTS (Bionic Beaver)</code></p>
<p>Kubernetes版本</p>
<p><code>Kubernetes v1.23.10</code></p>
<p>Karmada版本</p>
<p><code>Karmada v1.3.0-4-g1f13ad97</code></p>
<p>Karmada控制面所在的节点配置</p>
<ul>
<li class="">CPU</li>
</ul>
<div class="language-text codeBlockContainer_Ckt0 theme-code-block" style="--prism-color:#393A34;--prism-background-color:#f6f8fa"><div class="codeBlockContent_QJqH"><pre tabindex="0" class="prism-code language-text codeBlock_bY9V thin-scrollbar" style="color:#393A34;background-color:#f6f8fa"><code class="codeBlockLines_e6Vv"><span class="token-line" style="color:#393A34"><span class="token plain">Architecture:        x86_64</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">CPU op-mode(s):      32-bit, 64-bit</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">Byte Order:          Little Endian</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">CPU(s):              64</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">On-line CPU(s) list: 0-63</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">Thread(s) per core:  2</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">Core(s) per socket:  16</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">Socket(s):           2</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">NUMA node(s):        2</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">Vendor ID:           GenuineIntel</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">CPU family:          6</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">Model:               85</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">Model name:          Intel(R) Xeon(R) Gold 6266C CPU @ 3.00GHz</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">Stepping:            7</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">CPU MHz:             3000.000</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">BogoMIPS:            6000.00</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">Hypervisor vendor:   KVM</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">Virtualization type: full</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">L1d cache:           32K</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">L1i cache:           32K</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">L2 cache:            1024K</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">L3 cache:            30976K</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">NUMA node0 CPU(s):   0-31</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">NUMA node1 CPU(s):   32-63</span><br></span></code></pre></div></div>
<ul>
<li class="">内存</li>
</ul>
<div class="language-text codeBlockContainer_Ckt0 theme-code-block" style="--prism-color:#393A34;--prism-background-color:#f6f8fa"><div class="codeBlockContent_QJqH"><pre tabindex="0" class="prism-code language-text codeBlock_bY9V thin-scrollbar" style="color:#393A34;background-color:#f6f8fa"><code class="codeBlockLines_e6Vv"><span class="token-line" style="color:#393A34"><span class="token plain">Maximum Capacity: 512 GB</span><br></span></code></pre></div></div>
<ul>
<li class="">磁盘</li>
</ul>
<div class="language-text codeBlockContainer_Ckt0 theme-code-block" style="--prism-color:#393A34;--prism-background-color:#f6f8fa"><div class="codeBlockContent_QJqH"><pre tabindex="0" class="prism-code language-text codeBlock_bY9V thin-scrollbar" style="color:#393A34;background-color:#f6f8fa"><code class="codeBlockLines_e6Vv"><span class="token-line" style="color:#393A34"><span class="token plain">Disk /dev/vda: 200 GiB, 214748364800 bytes, 419430400 sectors</span><br></span></code></pre></div></div>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="组件参数配置">组件参数配置<a href="https://karmada.io/zh/blog/2022/10/26/test-report#%E7%BB%84%E4%BB%B6%E5%8F%82%E6%95%B0%E9%85%8D%E7%BD%AE" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<ol>
<li class="">karmada-apiserver</li>
</ol>
<div class="language-text codeBlockContainer_Ckt0 theme-code-block" style="--prism-color:#393A34;--prism-background-color:#f6f8fa"><div class="codeBlockContent_QJqH"><pre tabindex="0" class="prism-code language-text codeBlock_bY9V thin-scrollbar" style="color:#393A34;background-color:#f6f8fa"><code class="codeBlockLines_e6Vv"><span class="token-line" style="color:#393A34"><span class="token plain">--max-requests-inflight=2000</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">--max-mutating-requests-inflight=1000</span><br></span></code></pre></div></div>
<ol start="2">
<li class="">karmada-aggregated-server</li>
</ol>
<div class="language-text codeBlockContainer_Ckt0 theme-code-block" style="--prism-color:#393A34;--prism-background-color:#f6f8fa"><div class="codeBlockContent_QJqH"><pre tabindex="0" class="prism-code language-text codeBlock_bY9V thin-scrollbar" style="color:#393A34;background-color:#f6f8fa"><code class="codeBlockLines_e6Vv"><span class="token-line" style="color:#393A34"><span class="token plain">--kube-api-qps=200</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">--kube-api-burst=400</span><br></span></code></pre></div></div>
<ol start="3">
<li class="">karmada-scheduler</li>
</ol>
<div class="language-text codeBlockContainer_Ckt0 theme-code-block" style="--prism-color:#393A34;--prism-background-color:#f6f8fa"><div class="codeBlockContent_QJqH"><pre tabindex="0" class="prism-code language-text codeBlock_bY9V thin-scrollbar" style="color:#393A34;background-color:#f6f8fa"><code class="codeBlockLines_e6Vv"><span class="token-line" style="color:#393A34"><span class="token plain">--kube-api-qps=200</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">--kube-api-burst=400</span><br></span></code></pre></div></div>
<ol start="4">
<li class="">karmada-controller-manager</li>
</ol>
<div class="language-text codeBlockContainer_Ckt0 theme-code-block" style="--prism-color:#393A34;--prism-background-color:#f6f8fa"><div class="codeBlockContent_QJqH"><pre tabindex="0" class="prism-code language-text codeBlock_bY9V thin-scrollbar" style="color:#393A34;background-color:#f6f8fa"><code class="codeBlockLines_e6Vv"><span class="token-line" style="color:#393A34"><span class="token plain">--kube-api-qps=200</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">--kube-api-burst=400</span><br></span></code></pre></div></div>
<ol start="5">
<li class="">karmada-agent</li>
</ol>
<div class="language-text codeBlockContainer_Ckt0 theme-code-block" style="--prism-color:#393A34;--prism-background-color:#f6f8fa"><div class="codeBlockContent_QJqH"><pre tabindex="0" class="prism-code language-text codeBlock_bY9V thin-scrollbar" style="color:#393A34;background-color:#f6f8fa"><code class="codeBlockLines_e6Vv"><span class="token-line" style="color:#393A34"><span class="token plain">--kube-api-qps=40</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">--kube-api-burst=60</span><br></span></code></pre></div></div>
<ol start="6">
<li class="">karmada-etcd</li>
</ol>
<div class="language-text codeBlockContainer_Ckt0 theme-code-block" style="--prism-color:#393A34;--prism-background-color:#f6f8fa"><div class="codeBlockContent_QJqH"><pre tabindex="0" class="prism-code language-text codeBlock_bY9V thin-scrollbar" style="color:#393A34;background-color:#f6f8fa"><code class="codeBlockLines_e6Vv"><span class="token-line" style="color:#393A34"><span class="token plain">--quota-backend-bytes=8G</span><br></span></code></pre></div></div>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="测试执行">测试执行<a href="https://karmada.io/zh/blog/2022/10/26/test-report#%E6%B5%8B%E8%AF%95%E6%89%A7%E8%A1%8C" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p>在使用Clusterloader2进行性能测试之前，我们需要自己通过配置文件定义性能测试策略。 我们使用的配置文件如下：</p>
<details class="details_lb9f alert alert--info details_b_Ee" data-collapsed="true"><summary>unfold me to see the yaml</summary><div><div class="collapsibleContent_i85q"><div class="language-yaml codeBlockContainer_Ckt0 theme-code-block" style="--prism-color:#393A34;--prism-background-color:#f6f8fa"><div class="codeBlockContent_QJqH"><pre tabindex="0" class="prism-code language-yaml codeBlock_bY9V thin-scrollbar" style="color:#393A34;background-color:#f6f8fa"><code class="codeBlockLines_e6Vv"><span class="token-line" style="color:#393A34"><span class="token key atrule" style="color:#00a4db">name</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> test</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain" style="display:inline-block"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">namespace</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">   </span><span class="token key atrule" style="color:#00a4db">number</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token number" style="color:#36acaa">10</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain" style="display:inline-block"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">tuningSets</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">   </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">name</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> Uniformtinyqps</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">     </span><span class="token key atrule" style="color:#00a4db">qpsLoad</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">        </span><span class="token key atrule" style="color:#00a4db">qps</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token number" style="color:#36acaa">0.1</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">   </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">name</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> Uniform1qps</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">     </span><span class="token key atrule" style="color:#00a4db">qpsLoad</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">        </span><span class="token key atrule" style="color:#00a4db">qps</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token number" style="color:#36acaa">1</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain" style="display:inline-block"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">steps</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">   </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">name</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> Create deployment</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">     </span><span class="token key atrule" style="color:#00a4db">phases</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">        </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">namespaceRange</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">             </span><span class="token key atrule" style="color:#00a4db">min</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token number" style="color:#36acaa">1</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">             </span><span class="token key atrule" style="color:#00a4db">max</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token number" style="color:#36acaa">10</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">          </span><span class="token key atrule" style="color:#00a4db">replicasPerNamespace</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token number" style="color:#36acaa">20</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">          </span><span class="token key atrule" style="color:#00a4db">tuningSet</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> Uniformtinyqps</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">          </span><span class="token key atrule" style="color:#00a4db">objectBundle</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">             </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">basename</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> test</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">deployment</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">               </span><span class="token key atrule" style="color:#00a4db">objectTemplatePath</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token string" style="color:#e3116c">"deployment.yaml"</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">               </span><span class="token key atrule" style="color:#00a4db">templateFillMap</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">                  </span><span class="token key atrule" style="color:#00a4db">Replicas</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token number" style="color:#36acaa">1000</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">        </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">namespaceRange</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">             </span><span class="token key atrule" style="color:#00a4db">min</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token number" style="color:#36acaa">1</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">             </span><span class="token key atrule" style="color:#00a4db">max</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token number" style="color:#36acaa">10</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">          </span><span class="token key atrule" style="color:#00a4db">replicasPerNamespace</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token number" style="color:#36acaa">1</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">          </span><span class="token key atrule" style="color:#00a4db">tuningSet</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> Uniform1qps</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">          </span><span class="token key atrule" style="color:#00a4db">objectBundle</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">             </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">basename</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> test</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">policy</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">               </span><span class="token key atrule" style="color:#00a4db">objectTemplatePath</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token string" style="color:#e3116c">"policy.yaml"</span><br></span></code></pre></div></div><div class="language-yaml codeBlockContainer_Ckt0 theme-code-block" style="--prism-color:#393A34;--prism-background-color:#f6f8fa"><div class="codeBlockContent_QJqH"><pre tabindex="0" class="prism-code language-yaml codeBlock_bY9V thin-scrollbar" style="color:#393A34;background-color:#f6f8fa"><code class="codeBlockLines_e6Vv"><span class="token-line" style="color:#393A34"><span class="token comment" style="color:#999988;font-style:italic"># deployment.yaml</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">apiVersion</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> apps/v1</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">kind</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> Deployment</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">metadata</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">name</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token punctuation" style="color:#393A34">{</span><span class="token punctuation" style="color:#393A34">{</span><span class="token plain">.Name</span><span class="token punctuation" style="color:#393A34">}</span><span class="token punctuation" style="color:#393A34">}</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">labels</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">group</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> test</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">deployment</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">spec</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">replicas</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token punctuation" style="color:#393A34">{</span><span class="token punctuation" style="color:#393A34">{</span><span class="token plain">.Replicas</span><span class="token punctuation" style="color:#393A34">}</span><span class="token punctuation" style="color:#393A34">}</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">selector</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">matchLabels</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">app</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> fake</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">pod</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">template</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">metadata</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">labels</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">        </span><span class="token key atrule" style="color:#00a4db">app</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> fake</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">pod</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">spec</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">affinity</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">        </span><span class="token key atrule" style="color:#00a4db">nodeAffinity</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">          </span><span class="token key atrule" style="color:#00a4db">requiredDuringSchedulingIgnoredDuringExecution</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">            </span><span class="token key atrule" style="color:#00a4db">nodeSelectorTerms</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">              </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">matchExpressions</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">                  </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">key</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> type</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">                    </span><span class="token key atrule" style="color:#00a4db">operator</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> In</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">                    </span><span class="token key atrule" style="color:#00a4db">values</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">                      </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> fake</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">kubelet</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">tolerations</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">        </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">key</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token string" style="color:#e3116c">"fake-kubelet/provider"</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">          </span><span class="token key atrule" style="color:#00a4db">operator</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token string" style="color:#e3116c">"Exists"</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">          </span><span class="token key atrule" style="color:#00a4db">effect</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token string" style="color:#e3116c">"NoSchedule"</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">containers</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">        </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">image</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> fake</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">pod</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">          </span><span class="token key atrule" style="color:#00a4db">name</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token punctuation" style="color:#393A34">{</span><span class="token punctuation" style="color:#393A34">{</span><span class="token plain">.Name</span><span class="token punctuation" style="color:#393A34">}</span><span class="token punctuation" style="color:#393A34">}</span><br></span></code></pre></div></div><div class="language-yaml codeBlockContainer_Ckt0 theme-code-block" style="--prism-color:#393A34;--prism-background-color:#f6f8fa"><div class="codeBlockContent_QJqH"><pre tabindex="0" class="prism-code language-yaml codeBlock_bY9V thin-scrollbar" style="color:#393A34;background-color:#f6f8fa"><code class="codeBlockLines_e6Vv"><span class="token-line" style="color:#393A34"><span class="token comment" style="color:#999988;font-style:italic"># policy.yaml</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">apiVersion</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> policy.karmada.io/v1alpha1</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">kind</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> PropagationPolicy</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">metadata</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">name</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> test</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">spec</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">resourceSelectors</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">apiVersion</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> apps/v1</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">kind</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> Deployment</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">placement</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">replicaScheduling</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">replicaDivisionPreference</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> Weighted</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">      </span><span class="token key atrule" style="color:#00a4db">replicaSchedulingType</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> Divided</span><br></span></code></pre></div></div></div></div></details>
<p>Kubernetes资源详细的配置如下表所示：</p>
<table><thead><tr><th>Maximum Type</th><th>Maximum Value</th></tr></thead><tbody><tr><td>Number of Clusters</td><td>100</td></tr><tr><td>Number of Pods</td><td>2000000</td></tr><tr><td>Number of Nodes per Cluster</td><td>5000</td></tr><tr><td>Number of Pods per Cluster</td><td>20000</td></tr><tr><td>Number of Namespaces per Cluster</td><td>20</td></tr><tr><td>Number of Pods per Namespace</td><td>1000</td></tr></tbody></table>
<p>详细的测试方法和过程，可以参考</p>
<p><a href="https://github.com/kubernetes/perf-tests/blob/master/clusterloader2/docs/GETTING_STARTED.md" target="_blank" rel="noopener noreferrer" class="">https://github.com/kubernetes/perf-tests/blob/master/clusterloader2/docs/GETTING_STARTED.md</a></p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="测试结果">测试结果<a href="https://karmada.io/zh/blog/2022/10/26/test-report#%E6%B5%8B%E8%AF%95%E7%BB%93%E6%9E%9C" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p>APIResponsivenessPrometheus:</p>
<p><img decoding="async" loading="lazy" alt="readonly latency cluster" src="https://karmada.io/zh/assets/images/readonly_cluster-12dba6f4372bb4fd8b77e0e6224db417.png" width="1466" height="452" class="img_ev3q"></p>
<p><img decoding="async" loading="lazy" alt="readonly latency resource" src="https://karmada.io/zh/assets/images/readonly_resource-9ee45365e35af9102737603f4fef3433.png" width="1467" height="429" class="img_ev3q"></p>
<p><img decoding="async" loading="lazy" alt="readonly latency namespace" src="https://karmada.io/zh/assets/images/readonly_namespace-ef19a9b9d8a3743d86e279e42817269f.png" width="1466" height="429" class="img_ev3q"></p>
<p><img decoding="async" loading="lazy" alt="mutating latency" src="https://karmada.io/zh/assets/images/mutating-ecc4e348d7894ff5201b1613a0543e3a.png" width="1468" height="430" class="img_ev3q"></p>
<p><img decoding="async" loading="lazy" alt="readonly latency aggregated apiserver" src="https://karmada.io/zh/assets/images/readonly_aggregated_apiserver-833535447794fdcd2ca6a2e1c428c2c3.png" width="1465" height="429" class="img_ev3q"></p>
<p><img decoding="async" loading="lazy" alt="mutating latency aggregated apiserver" src="https://karmada.io/zh/assets/images/mutating_aggregated_apiserver-4147998b3ffb315177a48f34bf81cd2f.png" width="1465" height="429" class="img_ev3q"></p>
<p>Cluster Registration Latency:</p>
<table><thead><tr><th>metric</th><th>P50(ms)</th><th>P90(ms)</th><th>P99(ms)</th><th>SLO</th></tr></thead><tbody><tr><td>cluster_register</td><td>5356</td><td>6125</td><td>6904</td><td>N/A</td></tr></tbody></table>
<p><strong>Note</strong>: Karmada的Pull模式适合用于私有云的场景。与Push模式相比，成员集群会运行一个名为karmada-agent的组件。它会从控制面拉取用户提交的应用，并运行在成员集群中。
在Pull模式集群注册的过程中，它会包含安装karmada-agent的时间。如果karmada-agent的镜像已经准备完毕的话，它很大程度上取决于单集群内Pod启动的时延。
这里不过多讨论Pull模式的注册时延。</p>
<p>Resource Distribution Latency:</p>
<table><thead><tr><th>metric</th><th>P50(ms)</th><th>P90(ms)</th><th>P99(ms)</th><th>SLO</th></tr></thead><tbody><tr><td>cluster_schedule</td><td>12</td><td>15</td><td>32</td><td>N/A</td></tr><tr><td>resource_distribution(Push)</td><td>706</td><td>899</td><td>1298</td><td>2000</td></tr><tr><td>resource_distribution(Pull)</td><td>612</td><td>881</td><td>989</td><td>2000</td></tr></tbody></table>
<h3 class="anchor anchorTargetStickyNavbar_Vzrq" id="push-mode">Push Mode<a href="https://karmada.io/zh/blog/2022/10/26/test-report#push-mode" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h3>
<p>Etcd latency:</p>
<p><img decoding="async" loading="lazy" alt="etcd latency" src="https://karmada.io/zh/assets/images/push_etcd_latency-f3ae24b16fd88360c0aa7dd8256e51d8.png" width="910" height="292" class="img_ev3q"></p>
<p>Resource Usage:</p>
<p><img decoding="async" loading="lazy" alt="resource total" src="https://karmada.io/zh/assets/images/push_total-07d6cc75badbf1b6932d3a4557af8850.png" width="1831" height="295" class="img_ev3q"></p>
<p><img decoding="async" loading="lazy" alt="cpu" src="https://karmada.io/zh/assets/images/push_cpu-e782a7a815086af90c9ef2ea76d2e841.png" width="1829" height="257" class="img_ev3q"></p>
<p><img decoding="async" loading="lazy" alt="memory" src="https://karmada.io/zh/assets/images/push_memory-30b6696eec1591f6a59cb224b3f76eca.png" width="1829" height="258" class="img_ev3q"></p>
<h3 class="anchor anchorTargetStickyNavbar_Vzrq" id="pull-mode">Pull Mode<a href="https://karmada.io/zh/blog/2022/10/26/test-report#pull-mode" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h3>
<p>Etcd latency:</p>
<p><img decoding="async" loading="lazy" alt="etcd latency" src="https://karmada.io/zh/assets/images/pull_etcd_latency-225b4e5dcacc363843226f2e002d38e6.png" width="1466" height="661" class="img_ev3q"></p>
<p>Resource Usage:</p>
<p><img decoding="async" loading="lazy" alt="resource total" src="https://karmada.io/zh/assets/images/pull_total-13b82f9f7fafe49f5107cb97c503e95f.png" width="1830" height="294" class="img_ev3q"></p>
<p><img decoding="async" loading="lazy" alt="cpu" src="https://karmada.io/zh/assets/images/pull_cpu-3e848a94d91a6edeca135188d0d4b619.png" width="1828" height="258" class="img_ev3q"></p>
<p><img decoding="async" loading="lazy" alt="memory" src="https://karmada.io/zh/assets/images/pull_memory-eb55eac7332be7169904833f7bfaf17e.png" width="1829" height="256" class="img_ev3q"></p>
<p>成员集群中的karmada-agent消耗了40m CPU(cores)和266Mi Memory(bytes)。</p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="结论与分析">结论与分析<a href="https://karmada.io/zh/blog/2022/10/26/test-report#%E7%BB%93%E8%AE%BA%E4%B8%8E%E5%88%86%E6%9E%90" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p>在以上的测试结果中，API调用时延和资源分发时延均符合上述定义的SLIs/SLOs。在整个过程中，系统消耗的资源在一个可控制的范围。
因此，<strong>Karmada能稳定支撑100个大规模集群</strong>，并且管理超过500,000个节点和2,000,000个的pods。在生产中，Karmada能有效支持数以百计的大规模的集群。
接下来，我们会具体分析每个测试指标的数据。</p>
<h3 class="anchor anchorTargetStickyNavbar_Vzrq" id="关注点分离资源模板和策略">关注点分离：资源模板和策略<a href="https://karmada.io/zh/blog/2022/10/26/test-report#%E5%85%B3%E6%B3%A8%E7%82%B9%E5%88%86%E7%A6%BB%E8%B5%84%E6%BA%90%E6%A8%A1%E6%9D%BF%E5%92%8C%E7%AD%96%E7%95%A5" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h3>
<p>Karmada使用Kubernetes原生API来表达集群联邦资源模板，使用可复用的策略API来表达集群的调度策略。 它不仅可以让Karmada能够轻松集成Kubernetes的生态, 同时也大大减少了控制面的资源数量。
基于此，控制面的资源数量不取决于整个多集群系统集群的数量，而是取决于多集群应用的数量。</p>
<p>Karmada的架构集成了Kubernetes架构的简洁性和扩展性。Karmada-apiserver作为控制面的入口与Kubernetes的kube-apiserver类似。你可以使用单集群配置中所需的参数优化这些组件。</p>
<p>在整个资源分发过程中，API调用时延在一个合理的范围。</p>
<h3 class="anchor anchorTargetStickyNavbar_Vzrq" id="集群注册与资源分发">集群注册与资源分发<a href="https://karmada.io/zh/blog/2022/10/26/test-report#%E9%9B%86%E7%BE%A4%E6%B3%A8%E5%86%8C%E4%B8%8E%E8%B5%84%E6%BA%90%E5%88%86%E5%8F%91" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h3>
<p>在Karmada 1.3版本中，我们提供了基于Bootstrap tokens注册Pull模式集群的能力。这种方式不仅可以简化集群注册的流程，也增强了集群注册的安全性。
现在无论是Pull模式还是Push模式，我们都可以使用karmadactl工具来完成集群注册。与Push模式相比，Pull模式会在成员集群运行一个名为karmada-agent的组件。</p>
<p>集群注册时延包含了控制面收集成员集群状态所需的时间。在集群生命周期管理的过程中，Karmada会收集成员集群的版本，支持的API列表以及集群是否健康的状态信息。
此外，Karmada会收集成员集群的资源使用量，并基于此对成员集群进行建模，这样调度器可以更好地为应用选择目标集群。 在这种情况下，集群注册时延与集群的规模息息相关。
上述指标展示了加入一个5,000节点的集群直至它可用所需的时延。你可以通过<a href="https://karmada.io/docs/next/userguide/scheduling/cluster-resources#disable-cluster-resource-modeling" target="_blank" rel="noopener noreferrer" class="">关闭集群资源建模</a>来使集群注册时延与集群的大小无关，在这种情况下，集群注册时延这个指标将小于2s。</p>
<p>不论是Push模式还是Pull模式，Karmada都以一个很快的速度来下发资源到成员集群中。唯一的区别在于karmada-controller-manager负责所有Push模式集群的资源分发，而karmada-agent只负责它所在那一个Pull模式集群。
因此, 在高并发条件下发资源的过程中，Pull在相同配置条件下会比Push模式更快。
你也可以通过调整karmada-controller-manager的<code>--concurrent-work-syncs</code>的参数来调整同一时间段内并发work的数量来提升性能。</p>
<h3 class="anchor anchorTargetStickyNavbar_Vzrq" id="push模式和pull模式的资源使用量对比">Push模式和Pull模式的资源使用量对比<a href="https://karmada.io/zh/blog/2022/10/26/test-report#push%E6%A8%A1%E5%BC%8F%E5%92%8Cpull%E6%A8%A1%E5%BC%8F%E7%9A%84%E8%B5%84%E6%BA%90%E4%BD%BF%E7%94%A8%E9%87%8F%E5%AF%B9%E6%AF%94" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h3>
<p>在Karmada 1.3版本中，我们做了许多工作来减少Karmada管理大型集群的资源使用量。现在我们很高兴宣布，相比于1.2版本，Karmada 1.3在大规模场景下减少了 <strong>85%</strong> 的内存消耗和 <strong>32%</strong> 的CPU消耗。
总的来说, Pull模式在内存使用上有明显的优势，而在其他资源上相差的不大。</p>
<p>在Push模式中，控制面的资源消耗主要集中在karmada-controller-manager，而karmada-apiserver的压力不大。</p>
<p><img decoding="async" loading="lazy" alt="push apiserver qps" src="https://karmada.io/zh/assets/images/push_apiserver_qps-76f1db8c4eb993ab5803e0028702dfa1.png" width="908" height="291" class="img_ev3q"></p>
<p>从karmada-apiserver的qps以及karmada-etcd的请求时延我们可以看出karmada-apiserver的请求量保持在一个较低的水平。
在Push模式中，绝大多数的请求来自karmada-controller-manager。你可以配置<code>--kube-api-qps</code> and <code>--kube-api-burst</code>这两个参数来控制请求数在一个确定的范围内。</p>
<p>在Pull模式中，控制面的资源消耗主要集中在karmada-apiserver，而不是karmada-controller-manager。</p>
<p><img decoding="async" loading="lazy" alt="pull apiserver qps" src="https://karmada.io/zh/assets/images/pull_apiserver_qps-2c6f951f9d9e95fd8d97fe82d290899c.png" width="1466" height="432" class="img_ev3q"></p>
<p>从karmada-apiserver的qps以及karmada-etcd的请求时延我们可以看出karmada-apiserver的请求量保持在一个较高的水平。
在Pull模式中，每个成员集群的karmada-agent需要维持一个与karmada-apiserver通信的长连接。
我们很容易得出：在下发应用至所有集群的过程中karmada-apiserver的请求总量是是karmada-agent中配置的N倍(N=#Num of clusters)。
因此，在大规模Pull模式集群的场景下，我们建议增加karmada-apiserver的<code>--max-requests-inflight</code>以及<code>--max-mutating-requests-inflight</code>参数的值，和karmada-etcd的<code>--quota-backend-bytes</code>参数的值来提升控制面的吞吐量。</p>
<p>现在Karmada提供了<a href="https://karmada.io/docs/next/userguide/scheduling/cluster-resources" target="_blank" rel="noopener noreferrer" class="">集群资源模型</a>的能力来基于集群空闲资源做调度决策。
在资源建模的过程中，它会收集所有集群的节点与Pod的信息。这在大规模场景下会有一定的内存消耗。
如果你不使用这个能力，你可以<a href="https://karmada.io/docs/next/userguide/scheduling/cluster-resources#disable-cluster-resource-modeling" target="_blank" rel="noopener noreferrer" class="">关闭集群资源建模</a>来进一步减少资源消耗。</p>
<h3 class="anchor anchorTargetStickyNavbar_Vzrq" id="总结与展望">总结与展望<a href="https://karmada.io/zh/blog/2022/10/26/test-report#%E6%80%BB%E7%BB%93%E4%B8%8E%E5%B1%95%E6%9C%9B" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h3>
<p>根据测试结果分析，Karmada可以<strong>稳定支持100个大规模集群</strong>，管理超过50万个节点和200万个Pod。</p>
<p>在使用场景方面，Push模式适用于管理公有云上的Kubernetes集群，而Pull模式相对于Push模式涵盖了私有云和边缘相关的场景。在性能和安全性方面，Pull模式的整体性能要优于Push模式。
每个集群由集群中的karmada-agent组件管理，且完全隔离。
但是，Pull模式在提升性能的同时，也需要相应提升karmada-apiserver和karmada-etcd的性能，以应对大流量、高并发场景下的挑战。具体方法请参考kubernetes对大规模集群的优化。
一般来说，用户可以根据使用场景选择不同的部署模式，通过参数调优等手段来提升整个多集群系统的性能。</p>
<p>受限于测试环境和测试工具，本次测试尚未测试到Karmada多集群系统的上限，同时多集群系统的性能测试仍处于方兴未艾的阶段，下一步我们将继续优化多集群系统的测试工具，系统性地整理测试方法，以覆盖更大的规模和更多的场景。</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Karmada v1.3：更优雅 更精准 更高效]]></title>
            <link>https://karmada.io/zh/blog/2022/09/06/karmada-v1.3/karmada-v1.3</link>
            <guid>https://karmada.io/zh/blog/2022/09/06/karmada-v1.3/karmada-v1.3</guid>
            <pubDate>Tue, 06 Sep 2022 00:00:00 GMT</pubDate>
            <description><![CDATA[Karmada是开放的多云多集群容器编排引擎，旨在帮助用户在多云环境下部署和运维业务应用。凭借兼容Kubernetes原生API的能力，Karmada可以平滑迁移单集群工作负载，并且仍可保持与Kubernetes周边生态工具链协同。]]></description>
            <content:encoded><![CDATA[<p>Karmada是开放的多云多集群容器编排引擎，旨在帮助用户在多云环境下部署和运维业务应用。凭借兼容Kubernetes原生API的能力，Karmada可以平滑迁移单集群工作负载，并且仍可保持与Kubernetes周边生态工具链协同。</p>
<p>在最新发布的1.3版本中，Karmada重新设计了应用跨集群故障迁移功能，实现了基于污点的故障驱逐机制，并提供平滑的故障迁移过程，可以有效保障服务迁移过程的连续性（不断服）。</p>
<p>本版本新增加的特性：</p>
<ul>
<li class="">增加了面向多集群的资源代理新特性，通过该代理平台业务方可以在不感知多集群的情况下，以单集群访问姿势直接操纵部署在多集群的工作负载；</li>
<li class="">提供针对集群资源建模能力，通过自定义的集群资源模型，调度器可以更精准地进行资源调度；</li>
<li class="">提供基于Bootstrap令牌来注册Pull模式集群的能力，不仅可以简化集群注册过程，还可以方便地进行权限控制；</li>
</ul>
<p>此外，基于生产环境的用户反馈，本版本还进行了诸多性能优化，系统运行过程中CPU和内存资源需求大大降低，详细的性能测试报告稍后发布。</p>
<p>与之前版本一样，v1.3与前面的版本仍然保持兼容，前面版本的用户仍可以平滑升级。</p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="新特性概览">新特性概览<a href="https://karmada.io/zh/blog/2022/09/06/karmada-v1.3/karmada-v1.3#%E6%96%B0%E7%89%B9%E6%80%A7%E6%A6%82%E8%A7%88" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<h3 class="anchor anchorTargetStickyNavbar_Vzrq" id="基于污点的优雅驱逐">基于污点的优雅驱逐<a href="https://karmada.io/zh/blog/2022/09/06/karmada-v1.3/karmada-v1.3#%E5%9F%BA%E4%BA%8E%E6%B1%A1%E7%82%B9%E7%9A%84%E4%BC%98%E9%9B%85%E9%A9%B1%E9%80%90" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h3>
<p>当集群被判定为故障，并且故障时间超过宽限期（默认5分钟）之后，Karmada将为故障集群添加<code>NoExecute</code>污点，随后新引入的taint-manager控制器将开始驱逐该故障集群上的工作负载，接着调度器重新调度被驱逐的工作负载至新的可用集群，如果用户开启了GracefulEviction特性，被驱逐的工作负载并不会被立即删除，而是延迟到新的工作负载运行之后，可以保障驱逐过程中业务不中断。</p>
<p>整体故障迁移过程可以表示成："集群故障判定" → "负载预驱逐" → "重新调度" → "清理冗余负载"。</p>
<p>此处，无论故障判定还是驱逐，用户都可以参过参数来控制：</p>
<ul>
<li class=""><code>--failover-eviction-timeout</code>，指定从调度结果中删除故障集群的宽限期，默认5分钟</li>
<li class=""><code>--default-not-ready-toleration-seconds</code>，指定默认情况下添加到尚未具有<code>notReady:NoExecute</code>容忍的分发策略上的容忍时间，默认300秒</li>
<li class=""><code>--default-unreachable-toleration-seconds</code>，指定默认情况下添加到尚未具有<code>unreachable:NoExecute</code>容忍的分发策略上的容忍时间，默认300秒</li>
<li class=""><code>--graceful-eviction-timeout</code>，指定自工作负载已移动到优雅驱逐任务以来，等待优雅驱逐控制器执行最终删除的超时时间，默认时长10分钟</li>
</ul>
<h3 class="anchor anchorTargetStickyNavbar_Vzrq" id="跨多群集资源的全局代理">跨多群集资源的全局代理<a href="https://karmada.io/zh/blog/2022/09/06/karmada-v1.3/karmada-v1.3#%E8%B7%A8%E5%A4%9A%E7%BE%A4%E9%9B%86%E8%B5%84%E6%BA%90%E7%9A%84%E5%85%A8%E5%B1%80%E4%BB%A3%E7%90%86" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h3>
<p>Karmada在1.2版本中新增了一个<code>karmada-search</code>组件，该组件为选装组件，用于缓存集群中部署的资源对象和事件，并通过搜索API对外提供检索服务。</p>
<p>1.3版本中，我们在该组件中引入了一个新的代理功能，允许用户以访问单集群的方式访问多集群中的资源，无论资源是否由Karmada管理，利用代理功能，用户可以统一通过Karmada控制面来操作成员集群中的资源。</p>
<p>用户可以使用ResourceRegistry API来指定缓存的资源类型以及数据源（目标集群），例如以下配置表示从member1 和member2两个集群中缓存Pod与Node资源：</p>
<div class="language-yaml codeBlockContainer_Ckt0 theme-code-block" style="--prism-color:#393A34;--prism-background-color:#f6f8fa"><div class="codeBlockContent_QJqH"><pre tabindex="0" class="prism-code language-yaml codeBlock_bY9V thin-scrollbar" style="color:#393A34;background-color:#f6f8fa"><code class="codeBlockLines_e6Vv"><span class="token-line" style="color:#393A34"><span class="token key atrule" style="color:#00a4db">apiVersion</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> search.karmada.io/v1alpha1</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">kind</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> ResourceRegistry</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">metadata</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">name</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> proxy</span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain">sample</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token key atrule" style="color:#00a4db">spec</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">targetCluster</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">clusterNames</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> member1</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> member2</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">resourceSelectors</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">apiVersion</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> v1</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">kind</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> Pod</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">apiVersion</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> v1</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">kind</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> Node</span><br></span></code></pre></div></div>
<p>将该配置提交给karmada-apiserver之后，便可使用URL：<code>/apis/search.karmada.io/v1alpha1/proxying/karmada/proxy/api/v1/namespaces/default/pods</code>来进行集群资源访问。该URL中<code>/apis/search.karmada.io/v1alpha1/proxying/karmada/proxy</code>为固定前缀，后面部分与Kubernetes原生API路径完全一致。</p>
<p>关于该特性的更多信息可以参考：<a href="https://karmada.io/docs/userguide/globalview/proxy-global-resource/" target="_blank" rel="noopener noreferrer" class="">https://karmada.io/docs/userguide/globalview/proxy-global-resource/</a></p>
<h3 class="anchor anchorTargetStickyNavbar_Vzrq" id="基于自定义集群资源模型的调度">基于自定义集群资源模型的调度<a href="https://karmada.io/zh/blog/2022/09/06/karmada-v1.3/karmada-v1.3#%E5%9F%BA%E4%BA%8E%E8%87%AA%E5%AE%9A%E4%B9%89%E9%9B%86%E7%BE%A4%E8%B5%84%E6%BA%90%E6%A8%A1%E5%9E%8B%E7%9A%84%E8%B0%83%E5%BA%A6" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h3>
<p>在集群调度的过程中，karmada-scheduler会基于一系列的因素来做调度决策，其中一个不可或缺的因素就是集群的可用资源。之前的版本中，Karmada采用了一种通用的资源模型ResourceSummary来抽象集群的可用情况，如下所示：</p>
<div class="language-yaml codeBlockContainer_Ckt0 theme-code-block" style="--prism-color:#393A34;--prism-background-color:#f6f8fa"><div class="codeBlockContent_QJqH"><pre tabindex="0" class="prism-code language-yaml codeBlock_bY9V thin-scrollbar" style="color:#393A34;background-color:#f6f8fa"><code class="codeBlockLines_e6Vv"><span class="token-line" style="color:#393A34"><span class="token key atrule" style="color:#00a4db">resourceSummary</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">allocatable</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">cpu</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token string" style="color:#e3116c">"1024"</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">memory</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> 4096Mi</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">pods</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token string" style="color:#e3116c">"110"</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">allocated</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">cpu</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token string" style="color:#e3116c">"512"</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">memory</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> 2048Mi</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">pods</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token string" style="color:#e3116c">"64"</span><br></span></code></pre></div></div>
<p>但是ResourceSummary机械地累加了集群中所有节点的资源，忽视了节点上的碎片资源，这会导致资源需求较大的Pod无法准确地调度到合适的集群。同时它也忽视了不同用户的集群中节点可分配的资源不完全相同的特点。</p>
<p>1.3版本中Karmada引入了一种新的方式——自定义集群资源模型，来抽象集群的可用资源情况，旨在使调度器调度集群的结果更精确。用户可以启用--CustomizedClusterResourceModeling的特性开关来启用这一特性，开启后，在集群被Karmada所纳管后，Karmada会自动地为集群设置默认的资源模型，这一资源模型将集群中的各个节点分为不同等级的模型，默认的资源模型将根据CPU和内存这两项资源指标把节点分为9个不同的等级，如下所示：</p>
<div class="language-yaml codeBlockContainer_Ckt0 theme-code-block" style="--prism-color:#393A34;--prism-background-color:#f6f8fa"><div class="codeBlockContent_QJqH"><pre tabindex="0" class="prism-code language-yaml codeBlock_bY9V thin-scrollbar" style="color:#393A34;background-color:#f6f8fa"><code class="codeBlockLines_e6Vv"><span class="token-line" style="color:#393A34"><span class="token key atrule" style="color:#00a4db">resourceModels</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">grade</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token number" style="color:#36acaa">0</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">ranges</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">max</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token string" style="color:#e3116c">"1"</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">min</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token string" style="color:#e3116c">"0"</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">name</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> cpu</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">max</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> 4Gi</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">min</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token string" style="color:#e3116c">"0"</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">name</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> memory</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">grade</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token number" style="color:#36acaa">1</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">ranges</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">max</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token string" style="color:#e3116c">"2"</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">min</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token string" style="color:#e3116c">"1"</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">name</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> cpu</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">max</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> 16Gi</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">min</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> 4Gi</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">name</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> memory</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token punctuation" style="color:#393A34">...</span><span class="token plain">..</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"></span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">grade</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token number" style="color:#36acaa">8</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token key atrule" style="color:#00a4db">ranges</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">max</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token string" style="color:#e3116c">"9223372036854775807"</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">min</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token string" style="color:#e3116c">"128"</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">name</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> cpu</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">  </span><span class="token punctuation" style="color:#393A34">-</span><span class="token plain"> </span><span class="token key atrule" style="color:#00a4db">max</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> </span><span class="token string" style="color:#e3116c">"9223372036854775807"</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">min</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> 1Ti</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">    </span><span class="token key atrule" style="color:#00a4db">name</span><span class="token punctuation" style="color:#393A34">:</span><span class="token plain"> memory</span><br></span></code></pre></div></div>
<p>Cluster-status-controller将会收集集群内的节点、Pod信息计算对应模型节点的数量，与此同时，karmada-scheduler根据将要调度的实例资源请求比较不同集群中满足要求的节点数，并将实例调度到满足要求的节点更多的集群。</p>
<p>同时，在一些场景下，默认的集群资源模型不能满足用户特定集群的需求，现在用户可以通过kubectl edit cluster命令自定义地设置集群的资源模型，使资源模型能够更好地拟合集群的资源拓扑。</p>
<h3 class="anchor anchorTargetStickyNavbar_Vzrq" id="基于bootstrap令牌的集群注册">基于Bootstrap令牌的集群注册<a href="https://karmada.io/zh/blog/2022/09/06/karmada-v1.3/karmada-v1.3#%E5%9F%BA%E4%BA%8Ebootstrap%E4%BB%A4%E7%89%8C%E7%9A%84%E9%9B%86%E7%BE%A4%E6%B3%A8%E5%86%8C" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h3>
<p>1.3版本中，对于 Pull 模式下的集群，我们提供了一种通过命令行向 Karmada 控制面注册的方式。现在通过karmadactl token命令我们可以轻松的创建Bootstrap启动令牌的token，token的默认有效时长是24小时。</p>
<div class="language-shell codeBlockContainer_Ckt0 theme-code-block" style="--prism-color:#393A34;--prism-background-color:#f6f8fa"><div class="codeBlockContent_QJqH"><pre tabindex="0" class="prism-code language-shell codeBlock_bY9V thin-scrollbar" style="color:#393A34;background-color:#f6f8fa"><code class="codeBlockLines_e6Vv"><span class="token-line" style="color:#393A34"><span class="token plain">$ karmadactl token create --print-register-command </span><span class="token parameter variable" style="color:#36acaa">--kubeconfig</span><span class="token plain"> /etc/karmada/karmada-apiserver.config</span><br></span></code></pre></div></div>
<div class="language-shell codeBlockContainer_Ckt0 theme-code-block" style="--prism-color:#393A34;--prism-background-color:#f6f8fa"><div class="codeBlockContent_QJqH"><pre tabindex="0" class="prism-code language-shell codeBlock_bY9V thin-scrollbar" style="color:#393A34;background-color:#f6f8fa"><code class="codeBlockLines_e6Vv"><span class="token-line" style="color:#393A34"><span class="token comment" style="color:#999988;font-style:italic"># The example output is shown below</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain">karmadactl register </span><span class="token number" style="color:#36acaa">10.10</span><span class="token plain">.x.x:32443 </span><span class="token parameter variable" style="color:#36acaa">--token</span><span class="token plain"> t2jgtm.9nybj0526mjw1jbf --discovery-token-ca-cert-hash sha256:f5a5a43869bb44577dba582e794c3e3750f2050d62f1b1dc80fd3d6a371b6ed4</span><br></span></code></pre></div></div>
<p>通过karmadactl register命令可以在不复制成员集群kubeconfig的情况下非常轻松地完成包括部署karmada-agent在内的注册过程，增强了控制面以Pull模式纳管成员集群的易用性和安全性。</p>
<div class="language-shell codeBlockContainer_Ckt0 theme-code-block" style="--prism-color:#393A34;--prism-background-color:#f6f8fa"><div class="codeBlockContent_QJqH"><pre tabindex="0" class="prism-code language-shell codeBlock_bY9V thin-scrollbar" style="color:#393A34;background-color:#f6f8fa"><code class="codeBlockLines_e6Vv"><span class="token-line" style="color:#393A34"><span class="token plain">$ karmadactl register </span><span class="token number" style="color:#36acaa">10.10</span><span class="token plain">.x.x:32443 </span><span class="token parameter variable" style="color:#36acaa">--token</span><span class="token plain"> t2jgtm.9nybj0526mjw1jbf --discovery-token-ca-cert-hash sha256:f5a5a43869bb44577dba582e794c3e3750f2050d62f1b1dc80fd3d6a371b6ed4</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain" style="display:inline-block"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"> </span><span class="token comment" style="color:#999988;font-style:italic"># The example output is shown below</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"> </span><span class="token punctuation" style="color:#393A34">[</span><span class="token plain">preflight</span><span class="token punctuation" style="color:#393A34">]</span><span class="token plain"> Running pre-flight checks</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"> </span><span class="token punctuation" style="color:#393A34">[</span><span class="token plain">prefligt</span><span class="token punctuation" style="color:#393A34">]</span><span class="token plain"> All pre-flight checks were passed</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"> </span><span class="token punctuation" style="color:#393A34">[</span><span class="token plain">karmada-agent-start</span><span class="token punctuation" style="color:#393A34">]</span><span class="token plain"> Waiting to perform the TLS Bootstrap</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"> </span><span class="token punctuation" style="color:#393A34">[</span><span class="token plain">karmada-agent-start</span><span class="token punctuation" style="color:#393A34">]</span><span class="token plain"> Waiting to construct karmada-agent kubeconfig</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"> </span><span class="token punctuation" style="color:#393A34">[</span><span class="token plain">karmada-agent-start</span><span class="token punctuation" style="color:#393A34">]</span><span class="token plain"> Waiting the necessary secret and RBAC</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"> </span><span class="token punctuation" style="color:#393A34">[</span><span class="token plain">karmada-agent-start</span><span class="token punctuation" style="color:#393A34">]</span><span class="token plain"> Waiting karmada-agent Deployment</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"> W0825 </span><span class="token number" style="color:#36acaa">11</span><span class="token plain">:03:12.167027 </span><span class="token number" style="color:#36acaa">29336</span><span class="token plain"> check.go:52</span><span class="token punctuation" style="color:#393A34">]</span><span class="token plain"> pod: karmada-agent-5d659b4746-wn754 not ready. status: ContainerCreating</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"> </span><span class="token punctuation" style="color:#393A34">..</span><span class="token punctuation" style="color:#393A34">..</span><span class="token punctuation" style="color:#393A34">..</span><span class="token plain"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"> I0825 </span><span class="token number" style="color:#36acaa">11</span><span class="token plain">:04:06.174110 </span><span class="token number" style="color:#36acaa">29336</span><span class="token plain"> check.go:49</span><span class="token punctuation" style="color:#393A34">]</span><span class="token plain"> pod: karmada-agent-5d659b4746-wn754 is ready. status: Running</span><br></span><span class="token-line" style="color:#393A34"><span class="token plain" style="display:inline-block"></span><br></span><span class="token-line" style="color:#393A34"><span class="token plain"> cluster</span><span class="token punctuation" style="color:#393A34">(</span><span class="token plain">member3</span><span class="token punctuation" style="color:#393A34">)</span><span class="token plain"> is joined successfully</span><br></span></code></pre></div></div>
<h3 class="anchor anchorTargetStickyNavbar_Vzrq" id="版本升级">版本升级<a href="https://karmada.io/zh/blog/2022/09/06/karmada-v1.3/karmada-v1.3#%E7%89%88%E6%9C%AC%E5%8D%87%E7%BA%A7" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h3>
<p>我们验证了从Karmada 1.2版本到1.3版本的升级路径，升级过程平滑，可参考升级文档：<a href="https://karmada.io/docs/administrator/upgrading/v1.2-v1.3" target="_blank" rel="noopener noreferrer" class="">https://karmada.io/docs/administrator/upgrading/v1.2-v1.3</a></p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="致谢贡献者">致谢贡献者<a href="https://karmada.io/zh/blog/2022/09/06/karmada-v1.3/karmada-v1.3#%E8%87%B4%E8%B0%A2%E8%B4%A1%E7%8C%AE%E8%80%85" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<p>Karmada v1.3版本包含了来自51位贡献者的数百次代码提交，在此对各位贡献者表示由衷的感谢：</p>
<p>贡献者GitHub ID：</p>
<p>@AllenZMC
@calvin0327
@carlory
@CharlesQQ
@Charlie17Li
@chaunceyjiang
@cutezhangq
@dapengJacky
@dddddai
@duanmengkk
@Fish-pro
@Garrybest
@gy95
@halfrost
@hanweisen
@huntsman-li
@ikaven1024
@joengjyu
@JoshuaAndrew
@kerthcet
@kevin-wangzefeng
@kinzhi
@likakuli
@lonelyCZ
@luoMonkeyKing
@maoyangLiu
@mathlsj
@mikeshng
@Momeaking
@mrlihanbo
@my-git9
@nuclearwu
@Poor12
@prodanlabs
@RainbowMango
@suwliang3
@TheStylite
@wawa0210
@weilaaa
@windsonsea
@wlp1153468871
@wuyingjun-lucky
@XiShanYongYe-Chang
@xuqianjins
@xyz2277
@yusank
@yy158775
@zgfh
@zhixian82
@zhuwint
@zirain</p>
<h2 class="anchor anchorTargetStickyNavbar_Vzrq" id="参考链接">参考链接<a href="https://karmada.io/zh/blog/2022/09/06/karmada-v1.3/karmada-v1.3#%E5%8F%82%E8%80%83%E9%93%BE%E6%8E%A5" class="hash-link" aria-label="跳转到头部" title="跳转到头部" translate="no">​</a></h2>
<ul>
<li class="">Release Notes：<a href="https://github.com/karmada-io/karmada/releases/tag/v1.3.0" target="_blank" rel="noopener noreferrer" class="">https://github.com/karmada-io/karmada/releases/tag/v1.3.0</a></li>
<li class="">集群故障迁移使用指导：<a href="https://karmada.io/docs/userguide/failover/cluster-failover" target="_blank" rel="noopener noreferrer" class="">https://karmada.io/docs/userguide/failover/cluster-failover</a></li>
<li class="">多集群资源全局代理使用指导：<a href="https://karmada.io/docs/userguide/globalview/proxy-global-resource/" target="_blank" rel="noopener noreferrer" class="">https://karmada.io/docs/userguide/globalview/proxy-global-resource/</a></li>
<li class="">集群资源模型使用指导：<a href="https://karmada.io/docs/userguide/scheduling/cluster-resources" target="_blank" rel="noopener noreferrer" class="">https://karmada.io/docs/userguide/scheduling/cluster-resources</a></li>
<li class="">基于Bootstrap令牌的集群注册使用指导：<a href="https://karmada.io/docs/userguide/clustermanager/cluster-registration" target="_blank" rel="noopener noreferrer" class="">https://karmada.io/docs/userguide/clustermanager/cluster-registration</a></li>
</ul>]]></content:encoded>
        </item>
    </channel>
</rss>