org.apache.sling.discovery.oak.cluster

Class PartialStartupDetector

java.lang.Object

org.apache.sling.discovery.oak.cluster.PartialStartupDetector

public class PartialStartupDetector
extends Object

Discovery.oak requires that both Oak and Sling are operating normally in order to declare victory and announce a new topology.

The startup phase is especially tricky in this regard, since there are multiple elements that need to get updated (some are in the Oak layer, some in Sling):

• lease & clusterNodeId : this is maintained by Oak
• idMap : this is maintained by IdMapService
• leaderElectionId : this is maintained by OakViewChecker
• syncToken : this is maintained by SyncTokenService

A successful join of a cluster instance to the topology requires all 4 elements to be set (and maintained, in case of lease and syncToken) correctly.

This PartialStartupDetector is in charge of ensuring that a newly joined instance has all these elements set. Otherwise it is considered a "partially started instance" (PSI) and suppressed.

The suppression ensures that existing instances aren't blocked by a rogue, partially starting instance. However, there's also a timeout after which the suppression is no longer applied - at which point such a rogue instance will block existing instances. Infrastructure must ensure that a rogue instance is detected and restarted/fixed in a reasonable amount of time.

Method Summary

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait