[openstack-dev] [infra] placement job is busted in stable/newton (NO MORE HOSTS LEFT)

Matt Riedemann mriedem at linux.vnet.ibm.com
Thu Jan 12 02:40:26 UTC 2017


On 1/11/2017 8:18 PM, Matt Riedemann wrote:
> On 1/11/2017 9:19 AM, Jeremy Stanley wrote:
>>
>> If you look in the _zuul_ansible/scripts directory you'll see that
>> shell script which exited nonzero is the one calling devstack-gate,
>> so we've got something broken near the end of the job as you
>> surmise. I think it might be the post_test_hook:
>>
>> http://logs.openstack.org/57/416757/1/check/gate-tempest-dsvm-neutron-placement-full-ubuntu-xenial-nv/dfe0c38/logs/devstack-gate-post_test_hook.txt.gz
>>
>>
>> Looking in the nova repo, tools/hooks/post_test_hook.sh is a
>> relative symlink to gate/post_test_hook.sh but for some reason the
>> job doesn't seem to be following that. You might try recreating this
>> locally with the logs/reproduce.sh from that run and see if you get
>> the same behavior.
>>
>
> Hmm, I'm guessing this is somehow related to this:
>
> https://review.openstack.org/#/c/378952/
>
> But I'm not entirely sure how or why yet...I'll have to talk to Old Man
> Dague in the morning.
>

Well I guess it's less sinister than all that, it was just a matter of 
when the nova change landed, which was meant for newton but happened in 
Ocata:

https://review.openstack.org/#/c/376567/

So the script isn't there in newton. I'll push a change to the job in 
project-config to make that only run the hook if it exists.

-- 

Thanks,

Matt Riedemann




More information about the OpenStack-dev mailing list