[Gluster-infra] RPM build failures post-mortem

[Nigel Babu]

Hello folks,

You may have noticed that the RPM jobs failed quite often since the migration.
This should finally be fixed now. Here's a quick post-mortem of what happened.

After the restart, our Jenkins server has accidentally had a fr_FR locale. This
means some of the logs are now in French. In the past, we've had a permission
problem with the RPMS folder after a build. This is usually owned by root user
and root group. I fixed this by triggering a clean up after the job. The
trigger would listen for "Building remotely" in the log and then do a `chown`
on the workspace. Since we started Jenkins in the fr_FR locale, that line is
now in French, so Jenkins doesn't see "Bulding remotely" in the log. There's
now "Construction à distance" instead. This means the chown isn't triggered and
we error out with a Java traceback that's just a permission error. The Jenkins
user can't delete the workspace to clean it up.

The easy solution is to restart Jenkins. but in the interest of doing this
without a downtime, I've solved this with acls which create less of a mess.
This shouldn't be a problem from now on and I can take out the hacky
post-commit scripts.

Apologies for the large number of errors this week. If you recheck your job,
you should get a green now for rpm jobs.


--
nigelb
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 801 bytes
Desc: not available
URL: <http://lists.gluster.org/pipermail/gluster-infra/attachments/20170317/51056037/attachment.sig>