Stefan Egli created SLING-11496:
-----------------------------------
Summary: Fresh instance must remain suppressed until syncToken
stored
Key: SLING-11496
URL: https://issues.apache.org/jira/browse/SLING-11496
Project: Sling
Issue Type: Task
Components: Discovery
Reporter: Stefan Egli
Assignee: Stefan Egli
The changes in SLING-11450 have one case still missing : if an instance reuses
the clusterNodeId but is slow, it is not suppressed. Reason being that there's
no cleanup of data in /var/discovery/oak/idMap and ./clusterInstances. So if it
reuses the clusterNodeId, the old data from a previous instance would still be
there, and the other instances do not distinguish where the data originated.
The only way to detect a clusterNodeId-reuse is to require it to update the
syncToken. Until it doesn't do that it is suppressed. Once it does it, it joins
the cluster regularly. From then on, then syncToken is no longer checked (since
existing instances are excempted from that check).
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
