[chef] Re: Re: Application cookbook memory leak


Chronological Thread 
  • From: Kevin Nuckolls < >
  • To: chef < >
  • Subject: [chef] Re: Re: Application cookbook memory leak
  • Date: Fri, 23 Nov 2012 17:56:38 -0600

I can confirm seeing this in the wild on multiple nodes in our architecture. After upgrading everything to 10.16.2 and using the --fork flag for our chef-client daemon the problem has gone away. I had to merge in the --fork pull request myself, but it may have been merged into the chef-client cookbook repository by now. If it hasn't, it should. It's an important patch IMO.

https://github.com/knuckolls/chef-client/tree/tag-1.2.0-with-fork-patch

-Kevin


On Thu, Nov 22, 2012 at 2:46 PM, Arnold Krille < " target="_blank"> > wrote:
Hi,

On Thursday 22 November 2012 10:59:04 Karol Hosiawa wrote:
> I'm using
> https://github.com/opscode-cookbooks/application
> and
> https://github.com/opscode-cookbooks/application_ruby
> to deploy about 60 apps on different nodes
> running chef-client every 30 minutes as a service.
> After a few hours the chef-client process uses 500MB, after a day more
> than 1 GB.
> I have 14 nodes and only the ones that have the "application" cookbook
> in the run list leak memory.
<snip>
> I just wanted to ask if anyone else noticed it as well ? If no one else has
> then it suggests it's something specific to my setup and I should be looking
> for the leak somewhere else.

I can't help, I just noticed that I might have suffered the same problem. I
didn't investigate that deeply to pin it on the application-cookbook. But the
machine concerned has the application-cookbook applied and the others without
don't suffer mem-leaks.
I think I "solved" it by running chef-client from cron instead of as a deamon.
Gotta see how the machine behaves...

Have fun,

Arnold




Archive powered by MHonArc 2.6.16.

§