Details
-
Fix
-
Status: Dismissed (View Workflow)
-
Minor
-
Resolution: Works as designed
-
1.10.5
-
None
Description
As of now behavior
- Two distributed job chains are deployed on a 2 node JobScheduler active cluster.
- Each Job chains has only job e.g. job_chain1 has job1 and job_chain2 has job2.
- Both the jobs e.g. job1 and job2 share an exclusive lock test-exclusive which should restrict only 1 job to be executed at any given point of time, in the cluster.
- When both job chains are started one after another by submitting orders, job1 in job_chain1 will start and acquire the lock exclusively test-exclusive, for few second job2 in the job_chain2 will wait for lock and then it will also start on second active JobScheduler node.
- It seems that Locks has no effect in an active cluster architecture, because locks are managed locally.
How to reproduce
- Download attached file Ticket#2016082910000044.zip in a active JobScheduler cluster.
- First start job_chain1 and then job_chain2.
Desired behavior
- Locks should have same effect on an active cluster architecture as in a standalone JobScheduler.
- If two jobs share a exclusive lock, at any give time only one job should allowed to executed.
Maintainer Notes
- The Active Cluster is designed for distributed orders. This does not include distributed locks.
- The requirement for distributed locks will be discussed with JobScheduler Master branch 2.x
Attachments
Issue Links
- is related to
-
JITL-313 JITL JobSchedulerSynchronizeJobChains job should sync execution on an active cluster architecture
- Dismissed