MTMRIDUL TIWARIinmriduliti.hashnode.dev·Aug 1 · 9 min readEight Jenkins masters, one runbook, and the day AL2023 refused to behave like Amazon Linux 2The alert wasn't a pager — it was a 404 during dnf install jenkins on the first clone target. We were mid-wave on a Jenkins master migration: eight production controllers, each moving from an old host00
MTMRIDUL TIWARIinmriduliti.hashnode.dev·Jul 25 · 7 min readThe cron was fine — our log lifecycle was eating 51 GB of ghost diskThe third pager this week landed at 2:47 AM with the same subject line: root filesystem at 97%. I'd already verified the hourly S3 upload cron twice. sudo crontab -l showed the job firing at :30 every00
MTMRIDUL TIWARIinmriduliti.hashnode.dev·Jul 18 · 5 min readFrom 2 Minutes to 5 Seconds: Tracing Internal Domains with aws traceTask: Added trace domainTool: LazyOps extension — aws traceImpact: Domain-to-box lookup dropped from ~2 minutes to ~5 seconds https://pypi.org/project/lazyops-cli/ The Problem Nobody Talks About Inte00
MTMRIDUL TIWARIinmriduliti.hashnode.dev·Jul 12 · 7 min read97 evicted pods that refused to dieThe alert wasn't about user-facing errors. It was a node inventory check that came back wrong: 97 pods stuck in Evicted, all pinned to the same worker. Disk pressure had already done its job — the kub00
MTMRIDUL TIWARIinmriduliti.hashnode.dev·Jul 11 · 6 min readJenkins couldn't clone our repo — until we counted the SSH keysThe Jenkins job had been green for months. Then one morning it just… stopped. Same pipeline, same repo, same server — but the checkout stage hung until timeout. No fancy credential plugin, no stored s00