[
https://issues.apache.org/jira/browse/WICKET-5805?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14271390#comment-14271390
]
Martin Makundi commented on WICKET-5805:
----------------------------------------
1. Problem is not due to models, we use detachable models. Problem is due to
page serialisation, simply the component hierarchy takes too long to serialize
(couple of seconds is too long for user).
2. You can reproduce the session leak with the patch attached.
3. I will try to provide an example of (1.) where you can see the slowness.
> Non blocking page serialization (fixme: session leak)
> -----------------------------------------------------
>
> Key: WICKET-5805
> URL: https://issues.apache.org/jira/browse/WICKET-5805
> Project: Wicket
> Issue Type: Improvement
> Components: wicket
> Affects Versions: 1.4.23
> Reporter: Martin Makundi
> Priority: Critical
> Attachments: patch-serialize.txt
>
>
> We have a performance issue with our Wicket 1.4 app, page serialization
> causes inconvenience to user because RequestCycle.detach() blocks the request
> until session.requestDetached() has been handled.
> We attempted to solve this issue by invoking session.requestDetached in a
> separate thread and thus allowing user to receive their request without
> waiting for page serialization in SecondLevelCachePageMap. The disk writing
> is already parallel, but serialization is blocking.
> What would be the best way to implement this? I have attached our (proposed)
> patches.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
