Puppet users - Oct 2012 - Your number of resources and time of your catalog runs

Hi,

I was just wondering what is someone else''s number of manager resources
and
catalog run time.
Mine is 1,700 resources and average catalog run of 70-100 seconds (40 
seconds for catalog retrieval, 30-60 for catalog run), depending on the 
network distance between puppetmaster and managed node.
Then there are few more seconds for provider cache stuff, and report 
uploading takes it''s tool too, together adding some 20 seconds.

Care to share your puppet agent performance details?

b.

-- 
You received this message because you are subscribed to the Google Groups
"Puppet Users" group.
To view this discussion on the web visit
https://groups.google.com/d/msg/puppet-users/-/-BnVWSmxIIUJ.
To post to this group, send email to puppet-users@googlegroups.com.
To unsubscribe from this group, send email to
puppet-users+unsubscribe@googlegroups.com.
For more options, visit this group at
http://groups.google.com/group/puppet-users?hl=en.

On 10/22/2012 09:04 PM, Bostjan Skufca wrote:
> Hi,
> 
> I was just wondering what is someone else''s number of manager
resources
> and catalog run time.
> Mine is 1,700 resources and average catalog run of 70-100 seconds (40
> seconds for catalog retrieval, 30-60 for catalog run), depending on the
> network distance between puppetmaster and managed node.
> Then there are few more seconds for provider cache stuff, and report
> uploading takes it''s tool too, together adding some 20 seconds.
> 
> Care to share your puppet agent performance details?
We''re not as near in the number of resources per node yet, so no point
in comparing :) Let''s just say my client run is less then 10 seconds on
average :)


But as a side note, I''m wondering at what number of nodes one must
replace puppetmaster with Apache+Passanger? Just a rough estimate,
let''s
say master runs on 2 socket quadcore.

Is it 50 agents? 100 agents? 200 agents?

-- 
You received this message because you are subscribed to the Google Groups
"Puppet Users" group.
To post to this group, send email to puppet-users@googlegroups.com.
To unsubscribe from this group, send email to
puppet-users+unsubscribe@googlegroups.com.
For more options, visit this group at
http://groups.google.com/group/puppet-users?hl=en.

On Monday, October 22, 2012 4:12:15 PM UTC-5, Jakov Sosic
wrote:
>
> On 10/22/2012 09:04 PM, Bostjan Skufca wrote: 
> > Hi, 
> > 
> > I was just wondering what is someone else''s number of manager
resources
> > and catalog run time. 
> > Mine is 1,700 resources and average catalog run of 70-100 seconds (40 
> > seconds for catalog retrieval, 30-60 for catalog run), depending on
the
> > network distance between puppetmaster and managed node. 
> > Then there are few more seconds for provider cache stuff, and report 
> > uploading takes it''s tool too, together adding some 20
seconds.
> > 
> > Care to share your puppet agent performance details? 
>
> We''re not as near in the number of resources per node yet, so no
point
> in comparing :) Let''s just say my client run is less then 10
seconds on
> average :) 
>
>
> But as a side note, I''m wondering at what number of nodes one must
> replace puppetmaster with Apache+Passanger? Just a rough estimate,
let''s
> say master runs on 2 socket quadcore. 
>
>
I replaced it at around 35 nodes I think. I''ve currently got 42. Before
making the change to Apache + Passenger, we were seeing timeouts with 
webrick.

And in all honestly, I would make that change if you were using puppet in 
any sort of production environment. The only time I use webrick is for 
throwaway VMs where I have 1-3 nodes (including the master itself) running 
off a master so I can test something outside of my main master.

 
> Is it 50 agents? 100 agents? 200 agents? 
>
>
-- 
You received this message because you are subscribed to the Google Groups
"Puppet Users" group.
To view this discussion on the web visit
https://groups.google.com/d/msg/puppet-users/-/Uwq22fPQi5gJ.
To post to this group, send email to puppet-users@googlegroups.com.
To unsubscribe from this group, send email to
puppet-users+unsubscribe@googlegroups.com.
For more options, visit this group at
http://groups.google.com/group/puppet-users?hl=en.

On 10/22/2012 2:11 PM, Jakov Sosic wrote:
>
> But as a side note, I''m wondering at what number of nodes one must
> replace puppetmaster with Apache+Passanger? Just a rough estimate,
let''s
> say master runs on 2 socket quadcore.
>
> Is it 50 agents? 100 agents? 200 agents?
>
	Available CPU has no bearing on your question. Webrick is single 
threaded so the limiting factor is concurrency > 1. If your client 
requests aren''t overlapping then webrick will work fine. However if 
you''re using Puppet to support production servers I''d use 
Passenger/Apache regardless of whether you''re likely to hit the limits 
of webrick.

Ramin

-- 
You received this message because you are subscribed to the Google Groups
"Puppet Users" group.
To post to this group, send email to puppet-users@googlegroups.com.
To unsubscribe from this group, send email to
puppet-users+unsubscribe@googlegroups.com.
For more options, visit this group at
http://groups.google.com/group/puppet-users?hl=en.

Hi Jakov,

On 22 October 2012 23:11, Jakov Sosic wrote:
> We''re not as near in the number of resources per node yet, so no
point
> in comparing :) Let''s just say my client run is less then 10
seconds on
> average :)
>
Those were the nice old days, when puppet catalog was still small and it 
run fast :)


But as a side note, I''m wondering at what number of nodes one
must
> replace puppetmaster with Apache+Passanger? Just a rough estimate,
let''s
> say master runs on 2 socket quadcore.
>
> Is it 50 agents? 100 agents? 200 agents?
>
I''m thinking about it now, but at the distributed deployment, where
each
puppetmaster serves around 20-30 nodes it is really not worth a hassle by 
itself, but I will do it soon anyway, curiosity reasons. Also, nodes 
connect to master only once every hour, seems often enough.

b.

-- 
You received this message because you are subscribed to the Google Groups
"Puppet Users" group.
To view this discussion on the web visit
https://groups.google.com/d/msg/puppet-users/-/b2XfLOTIeFYJ.
To post to this group, send email to puppet-users@googlegroups.com.
To unsubscribe from this group, send email to
puppet-users+unsubscribe@googlegroups.com.
For more options, visit this group at
http://groups.google.com/group/puppet-users?hl=en.

On 10/22/2012 11:49 PM, Ramin K wrote:
> On 10/22/2012 2:11 PM, Jakov Sosic wrote:
>>
>> But as a side note, I''m wondering at what number of nodes one
must
>> replace puppetmaster with Apache+Passanger? Just a rough estimate,
let''s
>> say master runs on 2 socket quadcore.
>>
>> Is it 50 agents? 100 agents? 200 agents?
>>
> 
>     Available CPU has no bearing on your question. Webrick is single
> threaded so the limiting factor is concurrency > 1. If your client
> requests aren''t overlapping then webrick will work fine. However
if
> you''re using Puppet to support production servers I''d use
> Passenger/Apache regardless of whether you''re likely to hit the
limits
> of webrick.
> 
Thank you very much for your information, (both llowder and you).

I will migrate to Passanger ASAP. :)

-- 
You received this message because you are subscribed to the Google Groups
"Puppet Users" group.
To post to this group, send email to puppet-users@googlegroups.com.
To unsubscribe from this group, send email to
puppet-users+unsubscribe@googlegroups.com.
For more options, visit this group at
http://groups.google.com/group/puppet-users?hl=en.