[
https://issues.apache.org/jira/browse/OAK-7854?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16660333#comment-16660333
]
Michael Dürig commented on OAK-7854:
------------------------------------
Above patch does not account for flushes skipped because there were no changes.
This will make it difficult to define an alerting threshold based on the
monitoring endpoint. I therefore think we should monitor the attempted flushes
by the scheduler in the {{FileStore}}. If those stall there is a reason for an
alert and figuring out the root cause.
> Add liveliness monitoring for FileStore background operations
> ---------------------------------------------------------------
>
> Key: OAK-7854
> URL: https://issues.apache.org/jira/browse/OAK-7854
> Project: Jackrabbit Oak
> Issue Type: Improvement
> Components: segment-tar
> Reporter: Michael Dürig
> Assignee: Michael Dürig
> Priority: Major
> Fix For: 1.10
>
>
> The FileStore background operations are ultimately executed through a
> {{ScheduledExecutorService}}. In the case this scheduling gets blocked (e.g.
> because of a deadlock or lock contention in one of its tasks) there is chance
> of repository corruption.
> To minimise potential data loss we should implement monitoring endpoints for
> the vital background operations. This would allow deployments to take action
> early in case of failures and thus minimise potential data loss and simplify
> recovery.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
