Red Hat Bugzilla – Bug 997012
Reassignment of task does not happen after fail over
Last modified: 2014-08-06 16:11:53 EDT
Created attachment 786534 [details]
Description of problem:
See the attached process definition. It's a simple process containing a human task with notification & reassignment configured. This process has been deployed to jBPM cluster (two EAP 6.1 nodes) and started on the first node. The was waiting for the task to be started when node 1 was shut down. After waiting for 1 minute to let the cluster move the computation to the second node and the reassignment to take effect, the potential owner of the task hasn't changed. I would expect reassignment from "user" to "admin", but the potential owner was still "user".
I tried this scenario with standalone server (no cluster, no fail over) and it worked.
I am attaching server logs from both nodes, there are several stacktraces with quartz, so I guess they can be relevant to this issue.
The cluster has been started with clean DB (mysql 55) and erased .niogit on both nodes.
Version-Release number of selected component (if applicable):
Created attachment 786535 [details]
node one server log
Created attachment 786536 [details]
node two server log
problem was caused by the deadline job in cluster was not properly serialized and thus was not possible to be executed on another server than it was created. Mechanism for deadlines has been improved so it is properly serialized when storing the job details in db.
Fix requires commits from kie-commons and jbpm
Moving to an arbitrary future milestone, as this hasn't been fixed in ER1.
Verified on BPMS-6.0.0.ER4