Spin up a three-node Garage v2 S3 cluster on UnRAID + Ubuntu, add a slick WebUI, then wire Duplicacy to push encrypted, deduped backups—complete with cron jobs, Telegram alerts, and zone-aware triple replication—for a zero-friction, homelab-ready safety net.
Claude’s MCP lit a spark: in a weekend I taught GenieACS to talk to LLMs. Using Go, GoReleaser and GitOps, I exposed seven tools so an AI can list devices, reboot routers or flash firmware through a single /mcp endpoint.
Five years on-call taught me the pager’s first demand is context, not heroics. From half-built clusters to rogue upgrades, this post shares war stories, triage tactics, and manager tips for keeping incidents—and engineers—under control.
After almost decade in DevOps I spend as much time digging through forgotten repos as shipping code. Updating Sentry led me to GlitchTip, Redis Sentinel and hours of code-sleuthing—proof that documentation gaps turn DevOps into full-time software archaeologists.
Cluster idles at 8 % yet the bill soars? Learn how Prometheus data + KRR, a 50-line Python wrapper, Grafana and ArgoCD reclaimed 500 vCPU and 200 GiB across dozens of Kubernetes clusters—no magic, no incidents, just rightsizing done right.
