Process definition state in deployment cache can get out of sync in a multi-node cluster

[cnayak]

I am noticing an issue with Process Definitions that are activated/suspended on different nodes. It looks like the state of cached process definitions in the deployment cache can get out of sync. When they are out of sync, it impacts activation of process instances on the node that has a stale suspended state in the definition cache.

I had a suspended process definition which was re-activated subsequently on a node. An attempt to start process instances using the re-activated process definition fails on the second node but works on the first node that it was reactivated on. Debugging further, noticed that the state of the definition seemed to be out of sync on the two nodes.

Is this expected behavior or a known limitation ? If so, are there any workarounds to get them in sync ?

[yvoswillens]

Hi,

could you please create a JIRA with unit test so we can investigate this further?

Regards,

Yvo

[jbarrez]

Yes, i believe that suspension could indeed not work on multi node cluster.

Don't see quick workaround with the current version (except using a distributed cache .. which is kinda heavy)