Koji and mock init failure

[Mike McLean]

On 02/05/2010 05:17 PM, Michael Cronenworth wrote:
> I finally have my own Koji instance ironed out, using SSL instead of
> Kerberos.
>
> I was able to add a dist-f12 target and tags to Koji, and specified a
> localhost FTP site for a external repo. I submitted a custom SRPM I have
> as a scratch build to just test it out. The task resulted in a failure:
>
> BuildrootError: could not init mock buildroot, mock exited with status
> 20; see root.log for more information
>
> The root.log doesn't have any visible errors. I see it did a yum install
> of the base packages. I see that /var/lib/mock has two directories (for
> my two arches: i386, x86_64) and directories and binary files inside.
> Google didn't reveal any hints either. What am I missing?

Just a few notes about debugging mock failures. Hopefully, Dennis is 
right and all you need is to expand your build group. If not, maybe this 
will help.

Koji tries to guess which of the logs has the error. Sometimes it gets 
it wrong. If you don't see anything you should probably go ahead and 
check the other logs too.

Also, sometimes the error in the log may not jump out at you. If at 
first you don't find it, look again.

Sometimes you may need to replicate the mock error manually. Log into 
the build host and either user koji mock-config to clone the config (the 
--task option is handy for this situation). Optionally, as long as the 
failure was recent, you can just make a copy of the mock config that 
koji wrote for the task. It will be under /etc/mock/koji on the build 
host. Make a copy so that kojid doesn't delete it out from under you. 
You'll also want to edit config_opts['root'] in you copy so that you use 
a different chroot.

With some issues you may need to get yum to show more detail, you can 
set the debuglevel and/or errorlevel options higher in the 
config_opts['yum.conf'] setting of your mock config.