On Sun, Jan 23, 2011 at 1:59 PM, Daniel Pittman <dan...@puppetlabs.com> wrote:
> I would love to know what the details of your configuration are, including
> web server version (Apache, Passenger, etc), and what timeouts you have
> configured at that level.
Let's see here...
ii puppet 0.25.5-1gg11
centralised configuration management for networks
ii puppet-common 0.25.5-1gg11
common files for puppet and puppetmaster
ii puppetmaster 0.25.5-1gg11
centralised configuration management control daemon
(Our puppet .deb is built from the Debian 0.25.5 git tag. We have one
patch which removes the distribution config.ru from the package.)
These are all stock Ubuntu Lucid:
ii apache2-mpm-worker
2.2.14-5ubuntu8.4 Apache HTTP Server - high speed
threaded model
ii apache2-utils
2.2.14-5ubuntu8.4 utility programs for webservers
ii apache2.2-bin
2.2.14-5ubuntu8.4 Apache HTTP Server common binary
files
ii apache2.2-common
2.2.14-5ubuntu8.4 Apache HTTP Server common files
ii libapache2-mod-passenger 2.2.7debian-1
Rails and Rack support for Apache2
ii ruby1.8 1.8.7.249-2
Interpreter of object-oriented scripting language
Ruby 1
ii librack-ruby1.8 1.1.0-3
A modular Ruby webserver interface (Ruby 1.8)
Passenger config, which we've been using for awhile and I keep
forgetting to try and tune:
<IfModule mod_passenger.c>
PassengerMaxPoolSize 8
PassengerPoolIdleTime 300
PassengerStatThrottleRate 600
PassengerUseGlobalQueue on
PassengerHighPerformance on
</IfModule>
> I would also be interested to know what the concurrency levels you are
> seeing are like on the server, and ideally what sort of I/O you have - is
> this a heavily contended or slow disk, and how fast are writes generally
> completing?
Our config servers are all implemented as Xen virtual machines with 4
vCPUs and 4GB of RAM each. They are organized into VIPs with 2-4
backends per VIP. The VMs are dedicated to puppet but tend to share
physical hardware with VMs running over services. The occasional
spike aside, the typical config server VM under peak load exhibits:
- 15min load < 2.0
- CPU utilization < 20%
- No swapping
- < 15KB/s average disk throughput
- Write latency of 100-200ms or less
- 20-45 RESTful API calls per second (sorry, I don't have the detailed
breakdown of which API calls)
- < 8 catalog requests per minute across all environments
The service doesn't seem to be very loaded to me, especially compared
to the Bad Old Days where our config server VMs ran with 15min load
averages of 12-14 with the same hardware specs.
For the record, I can't actually upgrade to 2.6.x right now. My
natural inclination would have been to have done so prior to starting
this thread up in the first place but we have a defined type which
uses "stage" as a parameter. This conflicts hilariously with run
stages and I have to wait for an OS release with the fix to work its
way through the pipeline before I can resume work on a 2.6.x upgrade.
Jason
--
"His face disappeared. If someone has no face left, you know it's serious."
--
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+unsubscr...@googlegroups.com.
For more options, visit this group at
http://groups.google.com/group/puppet-users?hl=en.
