[
https://issues.apache.org/jira/browse/SLING-4048?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14174873#comment-14174873
]
Carsten Ziegeler commented on SLING-4048:
-----------------------------------------
I've finished a first version which is a proof of concept. In order to get this
done I changed a lot of other things on the fly as well.
Most notably, there is a new central TopologyHandler which keeps track of
changes in either the topology or the queue configuration and then internally
restarts the job handling.
With this change a lot of repository traversal is happening in parallel,
especially the clean up task and the jobs loader might traverse the tree
concurrently which leads to intermediate exceptions. In the end, this does not
matter that much but I think we should improve here.
> Avoid keeping jobs in memory
> ----------------------------
>
> Key: SLING-4048
> URL: https://issues.apache.org/jira/browse/SLING-4048
> Project: Sling
> Issue Type: Improvement
> Components: Extensions
> Reporter: Carsten Ziegeler
> Assignee: Carsten Ziegeler
> Fix For: Event 3.3.16
>
>
> Currently all jobs for a single instance are hold in memory and put into the
> queues. This is not optiomal, especially for a large amount of jobs. In
> addition it makes configuration changes, queue updates etc. more complicated.
> We should revert this and let a queue only pick a job from the resource tree
> when it has a free processing space.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
