On 12/30/2011 10:10 AM, 夜神 岩男 wrote:
On 12/30/2011 10:28 PM, Johnny Hughes wrote:
On 12/30/2011 04:55 AM, 夜神 岩男 wrote:
On 12/30/2011 01:45 PM, Karanbir Singh wrote:
my talk on application hosting on CentOS from Fosdem 2008 went into some details about this, there should be a copy of the talk online.
I appreciate the reference, I'll give it a look.
happy to answer specific questions... our build services are fairly robust as well ( remember over 100k builds were done leading to 6.0 ); the plague setup in place for c4/c5 has been fairly resilient as well.
Specific question. You mentioned plague. Are you using Koji now for build management or something else?
Something else ... no koji at all here.
...which would be?
Or was that supposed to be mysterious and intriguing?
No, It was supposed to answer the question that you asked. We are not using Koji ... we are using something else.
We have a custom build system. It uses a beanstalkd workqueue to submit packages to mock and can scale up or down as required and submit packages to several build machines. If we need to build 2500 packages, we can add extra machines to the system.
But mock on an x86_64 machine is what is actually building packages. The buildsystem around it could just as easily be the command line passing the proper variables to mock. In fact, it is just individual jobs being passed into mock and some other way to "queue" up the packages.