A Real Week, Not the YouTube Version

Monday: Reviewed 3 infrastructure PRs, triaged 2 new tickets from developers ("my pipeline is failing"), spent 2 hours debugging why a pod in staging was randomly OOMKilling at 3am (turned out to be a memory leak in a third-party library nobody noticed until it hit staging scale). Wrote the RCA. Updated the runbook.

Tuesday: Quarterly cloud cost review. Found ₹35K/month in idle resources. Opened tickets to clean them up. Had a 45-minute architecture discussion with the backend team about whether their new service needed its own RDS instance or could share the existing cluster. The answer was share — but only after a migration that took another hour to plan.

Wednesday: On-call. 2:47am — PagerDuty fires. One node in our EKS cluster is NotReady. Pods are being evicted. I'm awake in 4 minutes, logged in in 6. Node had hit disk pressure from a poorly configured log rotation job. Drained the node, cleaned up logs, brought it back, wrote the fix, deployed it. Back in bed by 4am. Tired but satisfied.

Thursday: All hands planning meeting. Estimated 3 sprints of infrastructure work for the new product launch. Had the "we need a staging environment that mirrors production" conversation for the 4th time with the same product manager. Agreed on a timeline. Spent the afternoon writing Terraform for the new staging VPC.

Friday: Dependency upgrades. Kubernetes cluster patching (1.29 → 1.30). Always stressful. Used kured for automatic node reboots. Watched metrics during the rolling upgrade. Nothing broke. Added 2 new monitoring dashboards. Closed 4 tickets. Left at 5:30pm feeling like a productive human being.

The Honest Breakdown of My Time

• 30% — Incident response and debugging (on-call weeks: higher)
• 25% — Planning, architecture discussions, meetings
• 20% — Building and improving infrastructure (the fun part)
• 15% — Code review, knowledge sharing, documentation
• 10% — On-call monitoring, cost optimization, maintenance