storm-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Stig Rohde Døssing (JIRA) <j...@apache.org>
Subject [jira] [Created] (STORM-3355) Make force kill delay for workers follow the supervisor's SUPERVISOR_WORKER_SHUTDOWN_SLEEP_SECS
Date Mon, 11 Mar 2019 19:32:00 GMT
Stig Rohde Døssing created STORM-3355:
-----------------------------------------

             Summary: Make force kill delay for workers follow the supervisor's SUPERVISOR_WORKER_SHUTDOWN_SLEEP_SECS
                 Key: STORM-3355
                 URL: https://issues.apache.org/jira/browse/STORM-3355
             Project: Apache Storm
          Issue Type: Improvement
    Affects Versions: 2.0.0
            Reporter: Stig Rohde Døssing


We currently have the supervisor.worker.shutdown.sleep.secs parameter allowing users to specify
how long the supervisor should wait between starting the initial graceful shutdown of a worker,
and sending the followup force kill. 

When workers are asked to shut down gracefully, they run a shutdown hook that allows 1 second
of cleanup, before force halting the JVM. I think it would be good to make the delay between
starting the shutdown hook and halting the JVM follow the same config as in the supervisor.


I don't see why it is useful to specify the force kill delay in the supervisor, if the worker
just suicides after one second anyway. Letting the user configure how long shutdown is allowed
to take lets them make use of the bolt's cleanup method for cleaning up resources in non-crash
scenarios.

Use case here https://stackoverflow.com/questions/55024919/resource-clean-up-after-killing-storm-topology



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

Mime
View raw message