<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Sre on Chetan&#39;s Portfolio</title>
    <link>https://chetan-thapliyal.cloud/tags/sre/</link>
    <description>Recent content in Sre on Chetan&#39;s Portfolio</description>
    <generator>Hugo -- gohugo.io</generator>
    <language>en</language>
    <copyright>© 2026 Chetan Thapliyal</copyright>
    <lastBuildDate>Fri, 20 Nov 2026 18:06:00 +0530</lastBuildDate><atom:link href="https://chetan-thapliyal.cloud/tags/sre/index.xml" rel="self" type="application/rss+xml" />
    
    <item>
      <title>Day 2 Before Day 2: Three Findings in Nine Minutes of K3s Upgrade</title>
      <link>https://chetan-thapliyal.cloud/posts/k3s-upgrade-day-2-runbooks/</link>
      <pubDate>Fri, 20 Nov 2026 18:05:00 +0530</pubDate>
      
      <guid>https://chetan-thapliyal.cloud/posts/k3s-upgrade-day-2-runbooks/</guid>
      <description>I wrote three runbooks for my homelab cluster before I needed them: predicting a drain, upgrading K3s by swapping one binary, and rotating the keys behind every secret. The next day the upgrade ran for real. About nine minutes of playbook found a command that had never worked, a node that evicted its own load-balancer agent, and pods that never moved back.</description>
      <media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://chetan-thapliyal.cloud/posts/k3s-upgrade-day-2-runbooks/featured.png" />
        <media:thumbnail xmlns:media="http://search.yahoo.com/mrss/" url="https://chetan-thapliyal.cloud/posts/k3s-upgrade-day-2-runbooks/featured.png" />
    </item>
    
    <item>
      <title>A Public Status Page That Can&#39;t Leak, and Can&#39;t Flatter Me Either</title>
      <link>https://chetan-thapliyal.cloud/posts/status-page-that-cannot-leak/</link>
      <pubDate>Fri, 20 Nov 2026 12:18:00 +0530</pubDate>
      
      <guid>https://chetan-thapliyal.cloud/posts/status-page-that-cannot-leak/</guid>
      <description>My homelab&amp;rsquo;s public status page shows recovery targets next to what a drill measured, incident response times computed from the reviews themselves, and a visit count with no tracking script, all under a strict Content Security Policy. How each part works, and why the page is built so it can&amp;rsquo;t make me look better than the evidence.</description>
      <media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://chetan-thapliyal.cloud/posts/status-page-that-cannot-leak/featured.png" />
        <media:thumbnail xmlns:media="http://search.yahoo.com/mrss/" url="https://chetan-thapliyal.cloud/posts/status-page-that-cannot-leak/featured.png" />
    </item>
    
    <item>
      <title>A Fix That Moved the Problem: The NAT That Only Let Packets Out, Twice</title>
      <link>https://chetan-thapliyal.cloud/posts/a-fix-that-moved-the-problem/</link>
      <pubDate>Fri, 20 Nov 2026 11:49:00 +0530</pubDate>
      
      <guid>https://chetan-thapliyal.cloud/posts/a-fix-that-moved-the-problem/</guid>
      <description>A worker&amp;rsquo;s tailnet path fell back to a relay, and moving it to a new UDP port fixed it. Eleven days later the ISP changed my home address and the new port broke the same way. Field notes on the alert the first review promised and nobody built, and how it works now.</description>
      <media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://chetan-thapliyal.cloud/posts/a-fix-that-moved-the-problem/featured.png" />
        <media:thumbnail xmlns:media="http://search.yahoo.com/mrss/" url="https://chetan-thapliyal.cloud/posts/a-fix-that-moved-the-problem/featured.png" />
    </item>
    
    <item>
      <title>The Rebuild I&#39;d Never Run: 48 Minutes Against a 15-Minute Target</title>
      <link>https://chetan-thapliyal.cloud/posts/homelab-dmz-observer-rebuild-drill/</link>
      <pubDate>Fri, 20 Nov 2026 11:47:58 +0530</pubDate>
      
      <guid>https://chetan-thapliyal.cloud/posts/homelab-dmz-observer-rebuild-drill/</guid>
      <description>I deleted my homelab&amp;rsquo;s outside observer and timed the documented rebuild for the first time. It took 48 minutes against a 15-minute target and hit three breaks that no earlier drill could reach, because each one lived in the machine&amp;rsquo;s identity. Plus the reverse case: VMs that decided they were new machines.</description>
      <media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://chetan-thapliyal.cloud/posts/homelab-dmz-observer-rebuild-drill/featured.png" />
        <media:thumbnail xmlns:media="http://search.yahoo.com/mrss/" url="https://chetan-thapliyal.cloud/posts/homelab-dmz-observer-rebuild-drill/featured.png" />
    </item>
    
    <item>
      <title>Disaster Recovery You Can Point To: What Each Drill Proved, and What It Didn&#39;t</title>
      <link>https://chetan-thapliyal.cloud/posts/homelab-disaster-recovery-drills/</link>
      <pubDate>Thu, 19 Nov 2026 11:06:47 +0530</pubDate>
      
      <guid>https://chetan-thapliyal.cloud/posts/homelab-disaster-recovery-drills/</guid>
      <description>Five written recovery plans for a GitOps homelab, scored by what was actually drilled: a cold restore from the offsite copy with the API up in 54 seconds, app restores checked row by row, an offsite copy that root can&amp;rsquo;t delete, and an honest list of what hasn&amp;rsquo;t been tested yet.</description>
      <media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://chetan-thapliyal.cloud/posts/homelab-disaster-recovery-drills/featured.png" />
        <media:thumbnail xmlns:media="http://search.yahoo.com/mrss/" url="https://chetan-thapliyal.cloud/posts/homelab-disaster-recovery-drills/featured.png" />
    </item>
    
    <item>
      <title>Sizing Kubernetes on a 12 GB Laptop: What the Scheduler Sees and What&#39;s Really There</title>
      <link>https://chetan-thapliyal.cloud/posts/k3s-memory-sizing-12gb-laptop/</link>
      <pubDate>Thu, 19 Nov 2026 11:00:41 +0530</pubDate>
      
      <guid>https://chetan-thapliyal.cloud/posts/k3s-memory-sizing-12gb-laptop/</guid>
      <description>Live numbers from a three-node K3s cluster on a 12 GB laptop: requests, limits and real usage per node, the control plane the scheduler can&amp;rsquo;t see, the limits that OOM kills corrected, and the gap I found while writing this (nothing reserved for the OS, no memory eviction at all) with the same-day fix.</description>
      <media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://chetan-thapliyal.cloud/posts/k3s-memory-sizing-12gb-laptop/featured.png" />
        <media:thumbnail xmlns:media="http://search.yahoo.com/mrss/" url="https://chetan-thapliyal.cloud/posts/k3s-memory-sizing-12gb-laptop/featured.png" />
    </item>
    
    <item>
      <title>Observability Off the Cluster: Prometheus and Loki in the Cloud, One Agent per Node</title>
      <link>https://chetan-thapliyal.cloud/posts/homelab-observability-off-cluster-loki-alloy/</link>
      <pubDate>Thu, 19 Nov 2026 10:29:27 +0530</pubDate>
      
      <guid>https://chetan-thapliyal.cloud/posts/homelab-observability-off-cluster-loki-alloy/</guid>
      <description>My homelab&amp;rsquo;s monitoring runs on a free cloud VM, not on the cluster it watches. Why the split, what each node runs, the memory budget that turned out wrong, swapping Promtail for Grafana Alloy without a gap, and the alerts that watch the monitoring itself.</description>
      <media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://chetan-thapliyal.cloud/posts/homelab-observability-off-cluster-loki-alloy/featured.png" />
        <media:thumbnail xmlns:media="http://search.yahoo.com/mrss/" url="https://chetan-thapliyal.cloud/posts/homelab-observability-off-cluster-loki-alloy/featured.png" />
    </item>
    
    <item>
      <title>Stopping an App Is a Change: The Cutover That Failed My Backup Heartbeat</title>
      <link>https://chetan-thapliyal.cloud/posts/stopping-an-app-is-a-change/</link>
      <pubDate>Thu, 19 Nov 2026 10:21:40 +0530</pubDate>
      
      <guid>https://chetan-thapliyal.cloud/posts/stopping-an-app-is-a-change/</guid>
      <description>Moving two apps to Kubernetes left their stopped databases in the old host&amp;rsquo;s backup script, and the nightly backup reported failure. Field notes on a small incident: why the alert was right, why it nearly went unnoticed, and the checklist line that fixes it.</description>
      <media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://chetan-thapliyal.cloud/posts/stopping-an-app-is-a-change/featured.png" />
        <media:thumbnail xmlns:media="http://search.yahoo.com/mrss/" url="https://chetan-thapliyal.cloud/posts/stopping-an-app-is-a-change/featured.png" />
    </item>
    
    <item>
      <title>The Power Cut That Outlasted the Inverter: A Homelab Cold Start, One Layer at a Time</title>
      <link>https://chetan-thapliyal.cloud/posts/homelab-power-cut-cold-start/</link>
      <pubDate>Thu, 19 Nov 2026 10:12:16 +0530</pubDate>
      
      <guid>https://chetan-thapliyal.cloud/posts/homelab-power-cut-cold-start/</guid>
      <description>A power cut drained the inverter and dropped my whole homelab mid-write. Getting it back took 15 hours and four separate failures, each hidden until the layer below it came up. The timeline, what held, and the fixes.</description>
      <media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://chetan-thapliyal.cloud/posts/homelab-power-cut-cold-start/featured.png" />
        <media:thumbnail xmlns:media="http://search.yahoo.com/mrss/" url="https://chetan-thapliyal.cloud/posts/homelab-power-cut-cold-start/featured.png" />
    </item>
    
    <item>
      <title>Seven Silent GitOps Gotchas: What ArgoCD and Helm Didn&#39;t Tell Me</title>
      <link>https://chetan-thapliyal.cloud/posts/seven-silent-gitops-gotchas-argocd-helm/</link>
      <pubDate>Mon, 16 Nov 2026 12:20:00 +0530</pubDate>
      
      <guid>https://chetan-thapliyal.cloud/posts/seven-silent-gitops-gotchas-argocd-helm/</guid>
      <description>Settings Helm accepted and ignored, apps that read Healthy while ArgoCD had stopped managing them, a webhook that refreshed everything at once, and four more ways a GitOps setup can look fine while it isn&amp;rsquo;t. Each one with its evidence and the check that catches it now.</description>
      <media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://chetan-thapliyal.cloud/posts/seven-silent-gitops-gotchas-argocd-helm/featured.png" />
        <media:thumbnail xmlns:media="http://search.yahoo.com/mrss/" url="https://chetan-thapliyal.cloud/posts/seven-silent-gitops-gotchas-argocd-helm/featured.png" />
    </item>
    
    <item>
      <title>Green Is Not a Status: Five Ways My Backups Reported Success While Failing</title>
      <link>https://chetan-thapliyal.cloud/posts/green-is-not-a-status/</link>
      <pubDate>Wed, 11 Nov 2026 23:00:00 +0530</pubDate>
      
      <guid>https://chetan-thapliyal.cloud/posts/green-is-not-a-status/</guid>
      <description>Every backup job in my homelab exited 0, and an audit found a different way each one was failing. Five silent successes, the evidence for each, and the guardrails that turned them into loud failures.</description>
      <media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://chetan-thapliyal.cloud/posts/green-is-not-a-status/featured.png" />
        <media:thumbnail xmlns:media="http://search.yahoo.com/mrss/" url="https://chetan-thapliyal.cloud/posts/green-is-not-a-status/featured.png" />
    </item>
    
  </channel>
</rss>
