Uploaded image for project: 'OpenIDM'
  1. OpenIDM
  2. OPENIDM-10061

Update clustered recon triggers to include state necessary to fix OPENIDM-8381



      https://bugster.forgerock.org/jira/browse/OPENIDM-8381 was filed because the recovery of scheduled jobs following cluster node failure did not work. The fix involved putting the field

      "recoverable" : "true"

      in the scheduled job as specified in https://stash.forgerock.org/projects/OPENIDM/repos/openidm/pull-requests/3476/overview

      This field should be added in the simple triggers used to run clustered recon pages by the SchedulerClusteredReconJobDispatch class - this field should be added to the JsonValue returned by SchedulerClusteredReconJobDispatch#getSchedulerDispatchState.

      Note that the fact that the recovery of scheduled jobs during cluster node failure did not work in the 5.5 time-frame, combined with complexities around distributed jobs and their recovery in the context of clustered-recon, meant that a clustered recon job will terminate when it detects that a cluster node went down. Thus the addition of the recoverable flag is not necessary in the 6.0 time-frame. It could be revisited in the 6.5 time-frame, but ONLY AFTER revisiting the complexities associated with allowing a clustered recon job to continue after a cluster node goes down. It is very possible that these issues are too onerous, and the best policy is the current policy of terminating a clustered recon when a cluster node goes down. In that case, the recoverable flag would not have to be added to the jobs scheduled as part of clustered recon.




            • Assignee:
              dhogan Dirk Hogan
              dhogan Dirk Hogan
            • Votes:
              0 Vote for this issue
              2 Start watching this issue


              • Created: