Class ClusterHealthRestored

java.lang.Object
org.keycloak.cluster.infinispan.ClusterHealthRestored
All Implemented Interfaces:
ProviderEvent

public class ClusterHealthRestored extends Object implements ProviderEvent
Published when KEYCLOAK_JDBC_PING2's periodic health check detects a transition from any non-healthy state back to HEALTHY. This covers:
  • Database reconnect — the health check could not reach the database (ERROR) and connectivity is restored.
  • Split-brain recovery — this node was in the losing partition (UNHEALTHY) and has rejoined the correct cluster view.
  • Coordinator election — no coordinator was present in the database table (NO_COORDINATOR) and one has since been elected.

Only available when the JDBC-PING cache stack is in use. Listened to by DatabaseAwareClusterProviderFactory in stateless mode, which broadcasts a CLEAR_ALL_LOCAL_CACHES_EVENT to all cluster nodes so that stale cache entries from the outage period are discarded.

  • Constructor Details

    • ClusterHealthRestored

      public ClusterHealthRestored()