<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>成本优化 on 黄文卓 | DevOps Engineer</title><link>https://socake.github.io/tags/%E6%88%90%E6%9C%AC%E4%BC%98%E5%8C%96/</link><description>Recent content in 成本优化 on 黄文卓 | DevOps Engineer</description><generator>Hugo -- gohugo.io</generator><language>zh-CN</language><managingEditor>17691281867@163.com (Wenzhuo Huang)</managingEditor><webMaster>17691281867@163.com (Wenzhuo Huang)</webMaster><copyright>© 2026 Wenzhuo Huang</copyright><lastBuildDate>Thu, 30 Apr 2026 16:00:00 +0800</lastBuildDate><atom:link href="https://socake.github.io/tags/%E6%88%90%E6%9C%AC%E4%BC%98%E5%8C%96/index.xml" rel="self" type="application/rss+xml"/><item><title>Playbook：K8s 成本优化实战——Karpenter + 弹性占位 + 精细 NodePool 的组合拳</title><link>https://socake.github.io/playbook/k8s-cost-optimization-karpenter/</link><pubDate>Thu, 30 Apr 2026 16:00:00 +0800</pubDate><author>17691281867@163.com (Wenzhuo Huang)</author><guid>https://socake.github.io/playbook/k8s-cost-optimization-karpenter/</guid><description>Karpenter 不是开箱即用的省钱按钮。把它跑出真实收益，需要先做 NodePool 按 workload 分层，再处理 sandbox/gpu 这类不被 K8s 识别的工作负载，最后用 placeholder 占位 Pod 弥合「扩容慢但缩容快」的体验缺口。本文给出可直接 kubectl apply 的完整 yaml 与可 chmod +x 直接跑的脚本，覆盖安装、四类 NodePool、弹性占位、S3 Gateway Endpoint、MQ 降级、监控与告警。</description><media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://socake.github.io/playbook/k8s-cost-optimization-karpenter/featured.jpg"/></item><item><title>Playbook：AWS MSK Serverless 迁回 Provisioned——什么时候、为什么、怎么迁</title><link>https://socake.github.io/playbook/msk-serverless-to-provisioned/</link><pubDate>Thu, 30 Apr 2026 13:30:00 +0800</pubDate><author>17691281867@163.com (Wenzhuo Huang)</author><guid>https://socake.github.io/playbook/msk-serverless-to-provisioned/</guid><description>MSK Serverless 看似按用量付费，实际上有一个常被忽视的最低消费层级：每个集群每月固定 $540 起、每个活跃消费者 IAM principal 还要按小时另收。对于流量长期 &amp;laquo; 1MB/s 的非生产环境，月费可以是同等吞吐 Provisioned 集群的 5-7 倍。本文记录将 4 个非生产环境从 MSK Serverless 迁回 Provisioned（kafka.t3.small × 2）的完整流程：成本计算脚本、aws kafka create-cluster 完整 JSON、IRSA 三 role 拆分、Java/Go/Python 三栈客户端配置、双集群双写五阶段切换、Schema Registry 导出导入、回滚脚本，以及踩过的多 IRSA、sarama、broker 数不可缩、Schema Registry 漏迁五个坑。</description><media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://socake.github.io/playbook/msk-serverless-to-provisioned/featured.jpg"/></item><item><title>Playbook：K8s 集群三合一实战——QA / PRE / AI Sandbox 合并的完整可执行手册</title><link>https://socake.github.io/playbook/k8s-cluster-consolidation/</link><pubDate>Thu, 30 Apr 2026 13:00:00 +0800</pubDate><author>17691281867@163.com (Wenzhuo Huang)</author><guid>https://socake.github.io/playbook/k8s-cluster-consolidation/</guid><description>集群合并的好处显性，坏处隐性。本 Playbook 不再停留在『讲个思路』，每段 yaml 都是完整 manifest（含 Namespace / ServiceAccount / RBAC / Secret），每段脚本都能 chmod +x 直接跑，每个步骤含前置 / 执行 / 验证 / 回滚四件套，并附一次真实事故的完整修复 SQL。</description><media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://socake.github.io/playbook/k8s-cluster-consolidation/featured.jpg"/></item><item><title>OpenCost 实战：Kubernetes 成本可见性与多团队费用分摊</title><link>https://socake.github.io/posts/opencost-kubernetes-cost-visibility/</link><pubDate>Sun, 12 Apr 2026 14:00:00 +0800</pubDate><author>17691281867@163.com (Wenzhuo Huang)</author><guid>https://socake.github.io/posts/opencost-kubernetes-cost-visibility/</guid><description>Kubernetes 成本不透明是 FinOps 落地的最大障碍。本文通过 OpenCost 构建完整的成本可见性体系，涵盖部署集成、云厂商价格接入、按团队分摊、Grafana 看板、超预算告警和自动周报推送，提供可直接复用的配置。</description><media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://socake.github.io/posts/opencost-kubernetes-cost-visibility/featured.jpg"/></item><item><title>FinOps 实践：Kubernetes 成本治理体系建设</title><link>https://socake.github.io/posts/finops-kubernetes-cost-governance/</link><pubDate>Sun, 12 Apr 2026 10:00:00 +0800</pubDate><author>17691281867@163.com (Wenzhuo Huang)</author><guid>https://socake.github.io/posts/finops-kubernetes-cost-governance/</guid><description>一套完整的 Kubernetes FinOps 落地路径：如何识别僵尸资源、配置成本分摊模型、利用 Karpenter 降低节点成本，以及如何将月账单从 $50k 压到 $30k。</description><media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://socake.github.io/posts/finops-kubernetes-cost-governance/featured.jpg"/></item><item><title>LLM 成本优化实战：从 Token 预算到模型路由</title><link>https://socake.github.io/posts/llm-cost-optimization/</link><pubDate>Mon, 19 Jan 2026 13:03:00 +0800</pubDate><author>17691281867@163.com (Wenzhuo Huang)</author><guid>https://socake.github.io/posts/llm-cost-optimization/</guid><description>我们的 AI 功能上线第一个月，LLM API 账单是 $18,000。通过模型路由、Prompt Caching 和 Batch API，第三个月降到了 $3,200。这篇文章记录具体怎么做到的。</description><media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://socake.github.io/posts/llm-cost-optimization/featured.jpg"/></item><item><title>Karpenter 弹性节点管理实战</title><link>https://socake.github.io/docs/kubernetes/karpenter-%E5%BC%B9%E6%80%A7%E8%8A%82%E7%82%B9/</link><pubDate>Mon, 08 Dec 2025 13:00:00 +0800</pubDate><author>17691281867@163.com (Wenzhuo Huang)</author><guid>https://socake.github.io/docs/kubernetes/karpenter-%E5%BC%B9%E6%80%A7%E8%8A%82%E7%82%B9/</guid><description>Karpenter 替代 Cluster Autoscaler 的完整实践：NodePool 约束配置、EC2NodeClass 实例选型、consolidation 节点整合降本、Spot 实例容错，以及多套集群配置的组织方式。</description></item><item><title>Kubernetes 成本优化实战：系统性降本的四条路径</title><link>https://socake.github.io/posts/k8s-%E6%88%90%E6%9C%AC%E4%BC%98%E5%8C%96%E5%AE%9E%E6%88%98/</link><pubDate>Mon, 18 Aug 2025 13:07:00 +0800</pubDate><author>17691281867@163.com (Wenzhuo Huang)</author><guid>https://socake.github.io/posts/k8s-%E6%88%90%E6%9C%AC%E4%BC%98%E5%8C%96%E5%AE%9E%E6%88%98/</guid><description>真实的降本案例：从发现成本异常到分析根因，通过 Karpenter 节点弹性伸缩、资源请求规格治理、大机型收敛等手段，系统性降低 AWS EC2 成本。包含具体配置和执行思路。</description><media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://socake.github.io/posts/k8s-%E6%88%90%E6%9C%AC%E4%BC%98%E5%8C%96%E5%AE%9E%E6%88%98/featured.jpg"/></item><item><title>Karpenter 深度解析：下一代 K8s 节点自动扩缩</title><link>https://socake.github.io/posts/karpenter-deep-dive/</link><pubDate>Wed, 11 Jun 2025 11:33:00 +0800</pubDate><author>17691281867@163.com (Wenzhuo Huang)</author><guid>https://socake.github.io/posts/karpenter-deep-dive/</guid><description>从 Cluster Autoscaler 迁移到 Karpenter 之后，集群扩容速度和节点利用率都有明显提升。本文详细拆解 Karpenter 的核心机制、关键配置项，以及在多套生产集群运行中踩过的坑。</description><media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://socake.github.io/posts/karpenter-deep-dive/featured.jpg"/></item><item><title>Descheduler 深度实战：Kubernetes 自动再平衡的正确打开方式</title><link>https://socake.github.io/posts/descheduler-workload-rebalance/</link><pubDate>Sat, 22 Mar 2025 16:00:00 +0800</pubDate><author>17691281867@163.com (Wenzhuo Huang)</author><guid>https://socake.github.io/posts/descheduler-workload-rebalance/</guid><description>kube-scheduler 只在 Pod 创建那一刻做决策，之后集群状态变了它就不管了。几个月下来，你的集群会变成 hot node + cold node 混杂、同一个 Deployment 的 Pod 全挤在一个 node、failure-domain 完全失衡。Descheduler 就是把调度决策后置、周期性重新评估的那只手。</description><media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://socake.github.io/posts/descheduler-workload-rebalance/featured.jpg"/></item></channel></rss>