Zeebe 0.23.0 - Delay in workflow task processing

I recently migrated from 0.22 to 0.23. Earlier the job task completes within a second, but now after migration its taking more time the same code and the same BPMN workflow.

On the client side am using spring-zeebe and in fact i added the requesttimeout and pollinterval parameters but no luck, earlier in 0.22 i was not having these parameters though.

    zeebe:
      client:
        broker:
          contactPoint: "zeebe-cluster-zeebe-gateway:26500"
        requestTimeout: "1s"
        job:
          pollInterval: "100ms"
        security:
          plaintext: true

Please let me know if am missing something ?Screen Shot 2020-05-07 at 9.37.28 AM

Hi @pumapathy, Can you put console statements in your code with timestamps to inspect where the time is being used.

@pumapathy also you might want to try 0.23.1

@salaboy Yes, I’m already using the 0.23.1 , my bad mentioned as 0.23 in the subject. BTW We are using the helm charts, please see below :

    image:
      repository: 10.56.249.91/camunda/zeebe
      tag: 0.23.1
      pullPolicy: IfNotPresent

@jwulf : Yes I had logger on the worker and time taken is very minimal ~200ms.

What I noticed, although the step got activated on the engine around 10:13:13 and the job poller is running on interval of every 100ms, the job doesn’t reach the worker until 10:13:22:187 and once handed off worker completes the jobs within ms. Please see the screenshot attached.

As I have mentioned earlier the same workflow doesn’t have this problem with 0.22 version , I’m thinking if there is any configuration being missed on the engine or something else ?

Screen Shot 2020-05-07 at 10.15.27 AM

@pumapathy that is interesting… is there anyway that you can create a reproducer so I can take a look?

@salaboy : Sure I have created an sample job, BPMN, HELM chart values used for engine .

Please let me know how do i attach the zip files , as alternate I have copied on the below location.

Please let me know if this is accessible.

What i observed is happening on the plain vanilla case is there some configuration introduced on the engine v0.23 which i missed or something else ?

Please see below the step activated on engine @ 12:04:48

Screen Shot 2020-05-11 at 12.05.31 PM

Although the job worker had poll interval of one second [tried in ms as well] the job was picked by poller until 12:05:02 and worker is just simple log stmts.

@pumapathy can you add the code in a GitHub repository, it is way much better to review stuff than a zip file that I need to download. Please share that GitHub repo when you have it.

@salaboy , Sure please see below the git details what was shared on the zip file. Could you please check and let me know ?