yellow team mailing list archive

[Graham Binns <graham@xxxxxxxxxxxxx>]

Further fun with the master:

On 13 February 2012 12:52, Gary Poster <gary.poster@xxxxxxxxxxxxx> wrote:
> We saw this hang before, and we thought we fixed it Friday. Francesco had a fix for the su contextmanager (a try/finally) and I landed it.
>
> The workaround I had before was to do a ps aux | grep python, and then look for a process that was trying to do a juju log as a buildbot user. That hangs, for some reason. Kill the process, and the parent script process will terminate, actually telling you what went wrong (in stdout or the logs) in a traceback.

Okay, so this problem was the result of me choosing the wrong fork to
go down with a conflict resolution. It's now fixed and I've pushed my
smart timeouts code into the master trunk.

However, if you try and run the tests on EC2, you'll see an error of
some sort, usually one of these two:

http://pastebin.ubuntu.com/840709/

or

http://pastebin.ubuntu.com/840713/

The second of these happens because after the first test is torn down,
the buildbot instance is still running on the machine against which
the tests are run. The next test encounters a start_error in the unit
as a result, because buildbot is already hogging port 8010.

I thought that we already killed running buildbot instances on
hooks/config-changed, but it seems like we don't always.

I haven't been able to work out the first one yet.

-- 
Graham Binns | PGP Key: EC66FA7D
http://launchpad.net/~gmb