[OpenStack-Infra] [openstack-infra][External CI][dsvm-tempest-full] Jenkins job failing at "Running devstack"

Jay Pipes jaypipes at gmail.com
Wed Sep 10 15:30:14 UTC 2014


On 09/10/2014 04:19 AM, Eduard Matei wrote:
> Hi guys,
> I'm working on setting up a CI testing platform for the Cinder driver
> we're developing.
> I followed
> http://www.joinfu.com/2014/02/setting-up-an-external-openstack-testing-system/
> and part 2, and i got both the master and the slave working, but the
> dsvm-tempest-full job is failing.
<snip>
> 2014-09-10 11:15:37.318 | error: The following untracked working tree files would be overwritten by checkout:
<snip>
> Any ideas how to fix this?

Hi Eduard,

I suspect, from looking at the above error, that you are re-using the 
slave instance on subsequent runs of Jenkins jobs. If you are doing this 
(as opposed to using something like nodepool to keep a pool of 
single-use slave VMs around for running Jenkins jobs, then you will need 
to run some sort of cleanup/reset script after the Jenkins job completes.

Akihiro Motoki has a nice repo with scripts you can use for this purpose:

https://github.com/amotoki/devstack-tools

(see the devstack-cleaner.sh script)

For best results, however, the upstream infrastructure team has 
definitely found that using single-use slave VMs, managed by nodepool, 
is the safest and cleanest way to handle Jenkins jobs that involve 
devstack-gate (like the jobs detailed in my joinfu.com posts).

All the best,
-jay



More information about the OpenStack-Infra mailing list