I have a cluster of 20 or so machines. On many of the nodes, the puppetd process take up a ton of resident memory (> 400MB RES in top, > 500GB VIRT in top) after running for a while (> 1 week). The only thread I found on this was titled "memory leak?" and last updated on 06-Dec-2007. The question on that tread was whether or not the memory was buffered memory. I can say that for my case, it is resident memory (RES under top). One other thing to note is that on nodes where the configuration is small (not many files synched), the memory doesn''t blow up that big (stays at < 100MB). Has anyone else seen this? Is there a ticket open for this? Env: I am running puppet 0.23.2 on CentOS 5.1. Using NFS to distribute files. -Chris _______________________________________________ Puppet-users mailing list Puppet-users@madstop.com https://mail.madstop.com/mailman/listinfo/puppet-users
Are you distributing large recursive trees of files? On 1/9/08, Chris Kline <chris@rapleaf.com> wrote:> > I have a cluster of 20 or so machines. On many of the nodes, the puppetd > process take up a ton of resident memory (> 400MB RES in top, > 500GB VIRT > in top) after running for a while (> 1 week). The only thread I found on > this was titled "memory leak?" and last updated on 06-Dec-2007. The > question on that tread was whether or not the memory was buffered memory. I > can say that for my case, it is resident memory (RES under top). One other > thing to note is that on nodes where the configuration is small (not many > files synched), the memory doesn''t blow up that big (stays at < 100MB). > > Has anyone else seen this? Is there a ticket open for this? > > > Env: I am running puppet 0.23.2 on CentOS 5.1. Using NFS to distribute > files. > > -Chris > > _______________________________________________ > Puppet-users mailing list > Puppet-users@madstop.com > https://mail.madstop.com/mailman/listinfo/puppet-users > >-- stickm@gmail.com -==< Stick >==- _______________________________________________ Puppet-users mailing list Puppet-users@madstop.com https://mail.madstop.com/mailman/listinfo/puppet-users
Indeed, although, I''ve noticed memory leaks in puppetmasterd and not
puppet. We''re currently restarting puppetmasterd via cron every 6
hours, to avoid file{} resources becoming corrupt in transit (instead of the
true contents, they contain an MD5 sum?)
As you can see, before Sunday at 12, puppetmasterd started using ridiculous
amounts of ram, and even started swapping. The box is currently only allocated
1GB of ram, and I believed this would be enough. (Luke, care to comment?)
http://homepages.maxnet.co.nz/djfu/puppet.maxnet.net.nz-memory-day.png
http://homepages.maxnet.co.nz/djfu/puppet.maxnet.net.nz-memory-week.png
Regards,
AJ (Fujin)
From: Chris MacLeod [mailto:stickm@gmail.com]
Sent: Saturday, 12 January 2008 5:34 a.m.
To: Puppet User Discussion
Subject: Re: [Puppet-users] memory leak in puppetd?
Are you distributing large recursive trees of files?
On 1/9/08, Chris Kline <chris@rapleaf.com> wrote:
I have a cluster of 20 or so machines. On many of the nodes, the puppetd
process take up a ton of resident memory (> 400MB RES in top, > 500GB VIRT
in top) after running for a while (> 1 week). The only thread I found on
this was titled "memory leak?" and last updated on 06-Dec-2007. The
question on that tread was whether or not the memory was buffered memory. I can
say that for my case, it is resident memory (RES under top). One other thing to
note is that on nodes where the configuration is small (not many files synched),
the memory doesn''t blow up that big (stays at < 100MB).
Has anyone else seen this? Is there a ticket open for this?
Env: I am running puppet 0.23.2 on CentOS 5.1. Using NFS to distribute files.
-Chris
_______________________________________________
Puppet-users mailing list
Puppet-users@madstop.com
https://mail.madstop.com/mailman/listinfo/puppet-users
--
stickm@gmail.com
-==< Stick >==-
On Jan 13, 2008, at 12:51 AM, Arjuna Christensen wrote:> Indeed, although, I''ve noticed memory leaks in puppetmasterd and not > puppet. We''re currently restarting puppetmasterd via cron every 6 > hours, to avoid file{} resources becoming corrupt in transit > (instead of the true contents, they contain an MD5 sum?) > > As you can see, before Sunday at 12, puppetmasterd started using > ridiculous amounts of ram, and even started swapping. The box is > currently only allocated 1GB of ram, and I believed this would be > enough. (Luke, care to comment?) > > http://homepages.maxnet.co.nz/djfu/puppet.maxnet.net.nz-memory-day.png > http://homepages.maxnet.co.nz/djfu/puppet.maxnet.net.nz-memory- > week.pngLooks like there''s a memory leak. Anyone able to, say, stick this under the dtrace microscope to figure out what''s going on? Or some other kind of microscope? In general, actually, anyone feel like spending some time doing analysis of this type on Puppet? I''ve done very little, and none in a long time. I''m now running Leopard, so I can actually use DTrace, but I''m low on time until February and I''m still trying to get 0.24.2 out ASAP. -- Finn''s Law: Uncertainty is the final test of innovation. --------------------------------------------------------------------- Luke Kanies | http://reductivelabs.com | http://madstop.com