Details
-
Improvement
-
Resolution: Unresolved
-
Major
-
None
-
18.8.0
-
None
-
Unknown
-
Description
DefaultRemoteObservationManager closes the JGroups channels as soon as ApplicationStoppedEvent is fired, before the component manager disposes anything. Every remote event produced during the rest of the shutdown (cache invalidations, job events) is lost for the other nodes.
Proposal: close the channels in JGroupsNetworkAdapter.dispose() with a dispose priority after the default components but before the stores, and drop incoming remote events while stopping. Also give the mail sender threads' join() a timeout so a stuck SMTP send cannot block the shutdown.
Attachments
Issue Links
- depends on
-
XCOMMONS-3824 Wait for running jobs when XWiki stops and mark interrupted jobs in their stored status
-
- Open
-
- is related to
-
XWIKI-25250 Events still queued when XWiki stops are lost because the remote Solr client is closed first
-
- Open
-
-
XWIKI-25256 Solr index updates still queued when XWiki stops are lost
-
- Open
-