[Shib-Users] IDP 2.3.0 memory hungry? Or something else in our setup?
Chad La Joie
lajoie at itumi.biz
Fri Aug 5 16:49:01 BST 2011
No, /proc/meminfo just uses the same underlying OS information as top
does. All that tells you is how much the memory the JVM has asked the
OS for, not how much of that is actually in use. You can plan the
upper limit the JVM will ever take by taking the Xmx setting and
adding about 120-150M to it to cover overhead of thread stack sizes,
perm gen, and just memory used by the JVM process directly when it
loads its executable.
On Fri, Aug 5, 2011 at 11:23, Wessel, Keith William
<kwessel at illinois.edu> wrote:
> Thanks, all. Yes, after Chad's advice against top, I've been avoiding that. But our system monitoring uses info from /proc/meminfo which I would think we can trust. As it's RHEL4, though, and thus has a backported patched 2.6.9 kernel, I'm sure it's not using the latest and greatest in memory management, thus the lingering unexplained memory usage after Tomcat and Terracotta are shut down.
>
> Long story short (and it's been a very long story), we're bumping our test VMs up to 4 GB to match production. That way, even the system monitoring tools should be happy. Then, we can (finally) move on with life.
>
> I thank everyone here for all the advice and input on this and, even though it's not a Shib-related problem that caused it, think we got some useful optimizing info out of the discussion.
>
> I now promise I'm going to put this issue to rest.
>
> Keith
>
>
> -----Original Message-----
> From: users-bounces at shibboleth.net [mailto:users-bounces at shibboleth.net] On Behalf Of Chad La Joie
> Sent: Thursday, August 04, 2011 7:06 PM
> To: Shib Users
> Subject: Re: [Shib-Users] IDP 2.3.0 memory hungry? Or something else in our setup?
>
> On Thu, Aug 4, 2011 at 19:16, Russell Beall <beall at usc.edu> wrote:
>> I'm not sure what the default PermGen size is, but Chad probably suggested removing it because the default is workable now. I saw 128M mentioned somewhere as the latest recommendation (when I looked a few months ago), so that is what I am planning to change it to. In production, mine is still set to 1G on the 2.1.5 release where the metadata loading hadn't yet been streamlined.
>
> Default perm gen is 64M on most JVMs on most systems. This is enough
> in most cases today. The tuning guide for the IdP recommends 128M
> just to catch the people with very large deployments.
>
> Note, 2.1.3 was the last time the especially high PermGen was really
> necessary. Prior to that there was a bug in Xerces that prevented
> loaded XML documents from going away if you pooled parsers and then a
> bug in the IdP that resulted in the same thing.
>
>> The histogram numbers you reported indicated that the actual Tomcat JVM had not filled up even remotely, so even though the system reported missing memory, it wasn't actually a problem. It is likely you do not have a problem with the actual JVMs.
>
> Yeah, I keep saying this over and over again. You *can not* use tools
> like top in order to gauge the amount of memory the JVM is actually
> using.
>
> --
> Chad La Joie
> www.itumi.biz
> trusted identities, delivered
> --
> To unsubscribe from this list send an email to users-unsubscribe at shibboleth.net
> --
> To unsubscribe from this list send an email to users-unsubscribe at shibboleth.net
>
--
Chad La Joie
www.itumi.biz
trusted identities, delivered
More information about the users
mailing list