[
https://issues.apache.org/jira/browse/OOZIE-2062?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14207704#comment-14207704
]
Shwetha G S commented on OOZIE-2062:
------------------------------------
Yes, we can delete old coord actions for long running coordinators. But
deleting coord action checks if the corresponding workflow is complete.
Deleting workflow checks if its sub-workflows are complete. All these checks
take time and overall purge takes longer. Instead, if we just delete all
completed workflows older than some time(without checking about sub-workflows)
and delete all coord actions older than some time(without checking their
corresponding workflows), purge is much faster. This is also a simpler logic to
maintain.
Both yahoo and inmobi have implemented this logic as an external cron. The idea
is to add it to oozie as an alternate purging logic
> Provide alternate purging logic
> -------------------------------
>
> Key: OOZIE-2062
> URL: https://issues.apache.org/jira/browse/OOZIE-2062
> Project: Oozie
> Issue Type: Improvement
> Reporter: Shwetha G S
> Assignee: Jaydeep Vishwakarma
>
> Currently, purge service deletes a workflow only if all its children are
> complete, coord action only if the corresponding workflow and its
> sub-workflows are complete. This logic is complicated and takes longer to
> run. Instead, provide an alternative purging logic that deleted all workflows
> whose end time is older than configured time, delete coord actions whose end
> time is older than configured time and so on. The purging logic to be used
> can be configured from oozie-site (by default can default to the current
> logic for backward compatbility)
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
