[chef] Re: Re: how to debug poor performance problems on Chef server?

["Julian C. Dunn" < >]

On Sat, Aug 25, 2012 at 1:25 AM, Jesse Nelson 
<
 >
 wrote:
> Most common issues i've seen that cause this behavior are:
> 1) Node object size ( if your using windows or some other ohai plugins
> that might populate the object with a lot of data). The requests can
> bog down the server.
>
> 2) Anti-pattern usage of node.save throughout cookbooks (which causes
> node to be committed back to server many times in a run)
>
> 3) Any combo of those and a tight run-interval. Or synchronized run
> interval with no splay.
>
> 4) Issues around couchdb ( but this doesn't seem like that from the
> tiny info you provided).
>
> You can run your chef-server with -N from the command line to see more
> about requests to your server.

Thanks for the info about "-N" and all the other suggestions.

I think I figured it out. We were using the lastrun report handler on
all nodes and for some reason it gets stuck in a state where it
hammers the Chef API, repeatedly updating the node information. Not
sure why it would do this, but the clue was that report handlers were
taking upwards of 15 minutes to run.

Once I disabled lastrun & restarted all my chef-clients, things
returned to normal. Might have something to do with the fact that I'm
on CentOS 5 and had to downgrade to Ruby 1.8.7 (long story)

- Julian