[
https://issues.apache.org/jira/browse/OAK-2624?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14379466#comment-14379466
]
Marcel Reutegger edited comment on OAK-2624 at 3/25/15 8:08 AM:
----------------------------------------------------------------
Thanks for the patch. Looks good to me.
I updated the patch slightly ([^OAK-2624-2.patch]) and changed some methods to
private, which are only used in DocumentNodeStore. I also removed the call to
{{internalRunBackgroundReadOperations()}} in {{dispose()}}. It is unnecessary
to pull in changes from other cluster nodes at this point. The changes will not
be processed anyway because the store is shutting down.
was (Author: mreutegg):
Thanks for the patch. Looks good to me.
I updated the patch slightly ([^OAK-2624-2.patch]) and changed some methods to
private, which are only used in DocumentNodeStore. I also removed the call to
{{internalRunBackgroundReadOperations()}} in {{dispose()}. It is unnecessary to
pull in changes from other cluster nodes at this point. The changes will not be
processed anyway because the store is shutting down.
> Move out non essential task processing from DocumentNodeStore background
> thread
> -------------------------------------------------------------------------------
>
> Key: OAK-2624
> URL: https://issues.apache.org/jira/browse/OAK-2624
> Project: Jackrabbit Oak
> Issue Type: Improvement
> Components: mongomk
> Reporter: Chetan Mehrotra
> Assignee: Stefan Egli
> Fix For: 1.1.8
>
> Attachments: OAK-2624-2.patch, OAK-2624.patch
>
>
> The {{DocumentNodeStore background thread}} handles quite a few task
> currently like
> # Clean orphaned branches
> # Clean collisions
> # Split documents
> # Background write - write back pending updates to _lastRev
> # Detect change in head revision from backend
> # Cache invalidation
> # Revision Comparator purge
> The background operation is scheduled to be performed every second (1s
> frequency). However at times some of the above task take more time which
> delay in picking up the changes from backend. For quick propagation of
> changes in cluster it is essential that work done in background thread is
> performed quickly. To enable that we should move out task which are not
> directly related to _head revision change_ like splitDocument.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
