<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>hpa on </title>
    <link>/tags/hpa/</link>
    <description>Recent content in hpa on </description>
    <generator>Hugo -- gohugo.io</generator>
    <language>en</language>
    <lastBuildDate>Sun, 12 Jul 2026 22:30:00 +0800</lastBuildDate><atom:link href="/tags/hpa/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>压测中副本数怎么都上不去:一次 ArgoCD selfHeal 与 HPA 抢 spec.replicas 的排查</title>
      <link>/posts/argocd-hpa-replicas-conflict/</link>
      <pubDate>Sun, 12 Jul 2026 22:30:00 +0800</pubDate>
      
      <guid>/posts/argocd-hpa-replicas-conflict/</guid>
      <description>背景 这个问题发生在 2026-07-09,我在一个 perf 环境里做容量压测,目标是找出交易链路上一个 Java 服务的吞吐上限。环境是共享的 EKS 集群,部署完全走 GitOps:ArgoCD 自动 sync,git 是唯一入口;被测服务开了 HPA。文中服务名统一替换为 pay-ms,环境、集群信息也做了模糊处理,配置片段保持原样只改名字。
被测服务在 Helm values 里的相关配置:
pay_ms: replicaCount: 3 resources: requests: { cpu: 4000m, memory: 4Gi } limits: { cpu: 4000m, memory: 4Gi } autoscaling: enabled: true minReplicas: 3 maxReplicas: 10 targetCPUUtilizationPercentage: 40 HPA 侧(kubectl describe hpa 摘要):
Min replicas: 3 Max replicas: 10 Metric: container CPU (pay-ms), target 40% of request Behavior: Scale Up: stabilization 60s; Max(Percent 100%/15s, Pods 4/15s) Scale Down: stabilization 600s; Pods 1/60s values 里同时写着 replicaCount: 3 和 autoscaling.</description>
    </item>
    
  </channel>
</rss>
