1

I'm using java quartz(2.3.1) and I have a setup where I'm using postgres as the job store and I have 3-4 machines all running the scheduler (vertical scaling). I want the rds to act as a source of truth and if I have a job with the schedule of repeat every 1 hour, I want it to run on any one of the machines. I don't care which one it runs on as long as it is one machine triggered in that hour.

I noticed that this works really well most of the time but I have recently had a trigger which runs once ever hour and about once every two days I see two of my machines getting triggered. I have noticed that my isClustered property is false which I have now set to true, but I'm not sure how this would help since if this was the problem, this issue would be happening a 100% of the time rather than rarely. Could anyone tell me what I should be looking into to actually fix this issue?

Soccergods
  • 440
  • 5
  • 17
  • 1
    This might help https://stackoverflow.com/questions/58513804/does-quartz-support-execute-once-on-multiple-node/58534137#58534137 .Also be sure to assign separate instanceId for all the clustered nodes. – user06062019 May 06 '20 at 16:00
  • Why is a separate instance ID something that's required? – Soccergods Aug 16 '20 at 10:42
  • 1
    Because clustering work on instance group and instance I'd.If you use the same instance id ,quartz scheduler will overwrite the entry made in the database from other instance with same I'd.Therefore, to overcome such scenerio using different or auto generated instance id is advisable.Note, quartz does clustering using database – user06062019 Aug 16 '20 at 12:31

1 Answers1

1

org.quartz.jobStore.isClustered = true

ensures proper database row locks are applied to the trigger before picking it, if that property was false both instances can pick up one trigger(race condition) before one could change status of that trigger.

http://www.quartz-scheduler.org/documentation/quartz-1.8.6/configuration/ConfigJDBCJobStoreClustering.html

vamsikurre
  • 309
  • 3
  • 5