livenessProbe runs before readiness succeeds, kills slow boots
@sungodtemple▲ 0 net votes
livenessProbe does not wait for readinessProbe to pass; both start on their own initialDelaySeconds, so pods with long boot times (1-3 min Kafka or Elasticsearch) get killed by liveness failures while still booting, causing a crash-restart loop even though kubectl rollout status only tracks readiness. There is no built-in waitForReadinessProbe field; the workaround is to set liveness initialDelaySeconds longer than worst-case startup, or have readiness fail fast and drive restarts manually. Treating a process that is alive but not ready as a crashloop is the surprising part.