On Mon, 2014-02-10 at 12:24 -0800, Andy Parker wrote:
> On Mon, Feb 10, 2014 at 12:07 PM, Brice Figureau
> <[email protected]> wrote:
>         On 10/02/14 20:46, Brice Figureau wrote:
>         > Hi Andy,
>         >
>         > This isn't directly related to this thread, but the current
>         master is
>         > almost 1.5 times slower (compilation, parsing wasn't taken
>         into
>         > consideration) than 3.4.2.
>         >
>         
>         > [snipped]
>         >
>         > I didn't bisect yet, just wanted your input if it's a known
>         regression,
>         > or if I should dig further.
>         
>         
>         The culprit seems to be:
>         
> https://github.com/puppetlabs/puppet/commit/9f00459daf809cfb782808076e61099d81ea535c
>         
>         I think the call to stale? might be the reason.
> 
> 
> Sounds likely. I tracked down bad performance to same code and
> submitted PR 2340 to address it. There may be side effects of my
> change, though. I'm a little worried that it might change how/when
> puppet loads manifests in a way that breaks someone's use.
> 
> 
> Can you try with PR 2340 applied and see if it helps?

I will make sure to test it and report success or failure.

> Also, I'd like to start putting together benchmarking scenarios so
> that we can all work from similar baselines of performance. Do you
> think there is some aspect of the way your puppet system works that we
> could make into a benchmark scenario and add to
> https://github.com/puppetlabs/puppet/tree/master/benchmarks ?

I need to study your benchmark system. I'm currently testing with my set
of production manifests, which I can't share publicly (it's about 1500
resources).
I could probably write a kind of generator that might model the
distribution of definition/resources/class/modules, but it will be kind
of hard to get the same level of resource relationship or even language
features used (or antipatterns).

Do you have a master <-> agent benchmarks machinery that I could reuse?

For the moment I was doing microbenchmarks of the agent, with a puppet
apply and debug logging. But I'd like to conduct agent tests as more
like in real life as possible...

Thanks,
-- 
Brice Figureau
My Blog: http://www.masterzen.fr/

-- 
You received this message because you are subscribed to the Google Groups 
"Puppet Developers" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To view this discussion on the web visit 
https://groups.google.com/d/msgid/puppet-dev/1392136509.421.19.camel%40arsenic.daysofwonder.com.
For more options, visit https://groups.google.com/groups/opt_out.

Reply via email to