You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@ignite.apache.org by "Gabriel Jimenez (JIRA)" <ji...@apache.org> on 2019/01/09 17:00:01 UTC

[jira] [Created] (IGNITE-10872) Queue Draining on Graceful Shutdowns

Gabriel Jimenez created IGNITE-10872:
----------------------------------------

             Summary: Queue Draining on Graceful Shutdowns
                 Key: IGNITE-10872
                 URL: https://issues.apache.org/jira/browse/IGNITE-10872
             Project: Ignite
          Issue Type: Improvement
          Components: general
            Reporter: Gabriel Jimenez


Our *issue to solve* was to have the grid complete any waiting and currently executing tasks on any termination where the signal was not KILL. With current behavior our tenants were losing data to corruption or incomplete requests due to job/task interruption on shutdown.

 Part of the functionality we were looking for is already exposed in the sense that the kill/stop functions within org.apache.ignite.Ignition already have the 'cancel' flag parameter. However this flag is hardcoded as 'true' in org.apache.ignite.internal.IgnitionEx's default shutdown hook (meaning jobs/tasks will be interrupted). Currently we have changed IgnitionEx's implementation to depend on an ignite system variable, and set default behavior to drain on termination with any signal other than KILL. Additional Questions: 

Was there an existing solution/approach to our '*issue to solve*' that did not involved changing the codebase? 

Is there another preferred solution for our issue to solve? 



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