Type: Regression Bug
Affects Version/s: 6.2.X EE, 7.0.X, Master
Component/s: Fault Tolerance > Clustering Framework
When redeploying first a delete gets done and then the job is scheduled again. Because from ClusterSchedulerEngine we trust master to notify slaves we don't do anything when is a MEMORY_CLUSTERED job in the slave. Hence, when the slave deploys after the master (or the master doesn't deploy) the job is deleted from the slave and not scheduled back
Steps to Reproduce:
- Deploy a cluster with a couple of nodes.
- Deploy a plugin wtih a defined job first in master, then in the slave.
- Check that the job is scheduled (com.liferay.portal.kernel.scheduler.SchedulerEngineHelperUtil.getScheduledJobs())
- Redeploy the plugin, first in master, then in the slave. The order is crucial.
- Stop the master.
- Check if the job is scheduled in the former slave (now master).
Job is deleted from _memoryClusteredJobs, so after step 6 job is no longer scheduled.
Job should remain in _memoryClusteredJobs, and after step 6 be scheduled.