[openstack-dev] [ironic] 100% failure in pxe_ssh job

Jim Rollenhagen jim at jimrollenhagen.com
Tue Sep 1 13:17:53 UTC 2015



> On Sep 1, 2015, at 04:32, Lucas Alvares Gomes <lucasagomes at gmail.com> wrote:
> 
> Hi,
> 
>> The current failure rate for the ironic pxe_ssh job is 100% -
>> http://graphite.openstack.org/render/?from=-200hours&height=500&until=now&width=800&bgcolor=ffffff&fgcolor=000000&yMax=100&yMin=0&target=color(alias(movingAverage(asPercent(stats.zuul.pipeline.check.job.gate-tempest-dsvm-ironic-pxe_ssh.FAILURE,sum(stats.zuul.pipeline.check.job.gate-tempest-dsvm-ironic-pxe_ssh.{SUCCESS,FAILURE})),%2712hours%27),%20%27gate-tempest-dsvm-ironic-pxe_ssh%27),%27orange%27)
>> 
>> The reason is something wrong with disk image builder and upstream ubuntu.
>> 
>> Which raises a much more pressing issue, why is an ironic integration
>> job building, from scratch a disk image builder image on every go (and
>> connecting to the internet to do it)? Especially as this job sits on a
>> bunch of other projects beyond ironic. Architecturally this is not sound
>> enough to be a voting job.
>> 
>> I'm proposing we make it non-voting immediately, and until it's redone
>> so it's no long dependent on pulling images directly from upstream, we
>> don't let it be voting.
> 
> Yeah, I had few time in the morning but I put a potential fix for that
> problem here to test if would solve the problem:
> https://review.openstack.org/#/c/219199/
> 
> This is one of the main jobs for Ironic and it would be great if we
> could keep it voting. Could we perhaps change the base OS to something
> else until Ubuntu is fixed ? ( Fedora / CentOS / Debian )

Why don't we just use the pre-built agent ramdisks we already publish? AFAIK we already have a job for it, just need to switch the name in project-config for Nova. 

// jim 




More information about the OpenStack-dev mailing list