Details
-
Fix
-
Status: Dismissed (View Workflow)
-
Blocker
-
Resolution: Won't Fix
-
1.9.11, 1.10.5
-
None
Description
Current situation
- Two distributed job chain with split and sync configuration deployed in an active JobScheduler cluster.
- If split job creates 2 order for subsequent parallel nodes, once split nodes executed the order, all the orders original order + 2 split order should be sync back at the sync node.
- In an active cluster JobScheduler nodes share the order execution load, which means any of the JobScheduler node can take over the queued Order and execute it, in some time the sync node expect 3 orders but get only 2 orders, which creates blockage.
- The missing 1 Order is executed on another active cluster member
- The exclusive lock across a cluster also does not stop the blockage, see
JS-1661
How to reproduce
- TBD
Desired behavior
- JITL JobSchedulerSynchronizeJobChains job should sync executions on an active JobScheduler cluster.
Developers Note
The sync job can not be implemented to be used in an active cluster. As a workaround you can use universal agents instead.
Attachments
Issue Links
- relates to
-
JS-1661 Exclusive lock.use has no effect in JobScheduler active cluster architecture
- Dismissed