Skip to content
gStack

/infra-runbook — incident runbook lookup

Symptom: $ARGUMENTS

Match the symptom to a runbook and present the diagnosis + the ordered, rehearsed fix, then wait for a go-ahead before any state-changing step.

  • PostgreSQL HA (invoke postgres-ha-ops): etcd NOSPACE (A) · divergent-timeline replica (B) · no-leader / DCS quorum loss (C) · node maintenance / switchover (D).
  • Kubernetes platform (invoke kubernetes-platform-ops): ArgoCD Unknown/ComparisonError (git creds, fix with prune off first) · pods Pending (no default SC) · node NotReady / read-only FS (disk full) · slow etcd / probe flapping · LoadBalancer pending (pool/overlap) · sealed Vault · Kyverno lockout.

Always: read-only diagnosis first, evidence before action, one change at a time, backup before destructive ops. Depth is in reference/ (the four Playbooks).