Skip to content
ReplicaDBReplicaDB

Operations troubleshooting

Start with /actuator/health/liveness and /actuator/health/readiness, then inspect logs and bounded metrics. A live process is not necessarily ready for database work.

  • Liveness up, readiness down: inspect PostgreSQL, Quartz, queue, or worker admission state before restarting.
  • Listener disconnected, polling healthy: notification latency may rise; polling preserves correctness. Check reconnect backoff and replicadb.worker.listener.connected. The listener retries from 1 second up to 30 seconds; inspect PostgreSQL LISTEN/NOTIFY, network, and firewall reachability while replicadb.managed.polling.lag shows fallback latency.
  • No claims: inspect queue age, replicadb.managed.claims, worker identity, permissions, available time, and lease expiry.
  • Repeated recovery: inspect lease renewal outcomes, database time, pool headroom, and retry backoff before adding workers.
  • Missing datasource values: verify the keyring version and do not replace a blank edit input with an empty security value.

Keep worker management endpoints private while collecting diagnostics.