Troubleshooting: High Availability¶

No fencing¶

Symptom: HA_NO_FENCING. Automatic failover is disabled because no fencing mechanism is configured. Existing servers keep running.

Configure fencing (power fencing / STONITH) before enabling automatic failover. Do not enable automatic recovery without fencing.

Quorum lost¶

Symptom: HA_QUORUM_LOST. Restore quorum before taking further action.

Split brain¶

Symptom: HA_SPLIT_BRAIN (CRITICAL). More than one node believes it owns a resource. Fence the stale owner manually. Do not auto-resolve; that can cause data loss.

Evacuation failed¶

Symptom: EVACUATION_FAILED. Review the evacuation items; some may be host-bound (GPU) and require manual recovery.

Recovery failed¶

Symptom: HA_RECOVERY_FAILED. Manual recovery is required.

Honesty reminder¶

HA moves are capability-gated but hardware-unproven in this build.