[Deploying Sakai] Measuring session size

[Sam Ottenhoff]

Getting a live copy of a heap dump during term time is really the best way
to do it in my experience.  Using gdb makes this really easy:
https://blogs.atlassian.com/2013/03/so-you-want-your-jvms-heap/

Aaron is the expert right now in Sakai sessions.  I think he's diagnosed a
fair number of horrible ways we are abusing our session with stuff that
shouldn't be there.

For JSF, I believe SAK-25677 is a pretty huge win and John Bush's graphs
are pretty compelling.  We are also running it in production for one client
and will start rolling it out for more clients.

Here is another ticket that has some interesting back and forth re assumed
memory use in a cache: https://jira.sakaiproject.org/browse/KNL-1056


On Tue, Mar 25, 2014 at 5:51 AM, Matthew Buckett <
matthew.buckett at it.ox.ac.uk> wrote:

> Does anyone have any good idea about Sakai session size on the heap.
> We're considering increasing our session timeout from 2 hours but
> would like to have an idea about the impact of this on our heap usage.
> I've had a quick go at using sizeof
> (https://github.com/dweiss/java-sizeof) but it seems that our object
> tree is huge, most likely because something in the session holds a
> reference to a spring service, and once you've got a reference to one
> of them you can walk the object graph to almost all the others. Better
> sizeof allows filtering of the objects you walk but I couldn't manage
> to get this to easily filter out all the services from the objects
> walked.
>
> Looking at a heap dump would give us an idea but we our currently out
> of term time so it may not be a very good indicator.
>
>
> Side note:
> - There's been some good discussion about session size on:
> https://jira.sakaiproject.org/browse/SAK-25677
>
> --
>   Matthew Buckett, VLE Developer, IT Services, University of Oxford
> _______________________________________________
> production mailing list
> production at collab.sakaiproject.org
> http://collab.sakaiproject.org/mailman/listinfo/production
>
> TO UNSUBSCRIBE: send email to
> production-unsubscribe at collab.sakaiproject.org with a subject of
> "unsubscribe"
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://collab.sakaiproject.org/pipermail/production/attachments/20140325/1b868929/attachment.html