[
https://issues.apache.org/jira/browse/SLING-3399?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13912681#comment-13912681
]
Michael Dürig commented on SLING-3399:
--------------------------------------
As discussed with [~cziegeler], we might be better of in investing into a new
solution targeted at Oak's scalability level (i.e. SLING-3279) instead of
tweaking {{JcrResourceListener}}, which would only push the bottleneck to the
next level.
Such a new solution would probably not be fully backward compatible with
{{JcrResourceListener}} but would scale much better in the face of many events.
Users would then actually have the choice between the two options trading off
full backwards compatibility against better scalability.
> OOME in JcrResourceListener for many events
> -------------------------------------------
>
> Key: SLING-3399
> URL: https://issues.apache.org/jira/browse/SLING-3399
> Project: Sling
> Issue Type: Bug
> Components: JCR
> Affects Versions: JCR Resource 2.3.0
> Reporter: Michael Dürig
>
> Oak supports large number of transient changes to be saved with a single save
> call. As a consequence such an operation will generate a big batch of
> observation events, which can cause the JcrResourceListener to run out of
> heap space.
> See also SLING-3279 where I already noted that queueing observation events
> might be problematic.
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)
