[openstack-dev] [Neutron] Fix migration that breaks Grenade jobs

Salvatore Orlando sorlando at nicira.com
Tue Apr 22 12:38:14 UTC 2014


When I initially spoke to the infra team regarding this problem, they
suggested that "just fixing migrations" so that the job could run was not a
real option.
I tend to agree with this statement.
However, I'm open to options for getting grenade going until the migration
problem is solved. Ugly workarounds might be fine, as long as we won't be
doing anything that a real deployer would never do.

Personally, I still think the best way for getting grenade to work is to
ensure previos_rev and current_rev have the same configuration. For the
havana/icehouse upgrade, this will mean that devstack for icehouse should
not add the metering plugin. I and Jakub are overdue a discussion on
whether this would be feasible or not.

Change 87935 is acceptable as a fix for that specific migration.
However it does not fix the general issue, where the root cause is that
currently the state of the neutron database depends on configuration
settings, and therefore migrations are idempotent as long as the plugin
configuration is not changed, which is not the case.

Salvatore


On 22 April 2014 11:14, Jakub Libosvar <libosvar at redhat.com> wrote:

> On 04/22/2014 10:53 AM, Anna Kamyshnikova wrote:
> > Hello everyone!
> >
> > I'm working on fixing bug 1307344. I found out solution that will fix
> > Grenade jobs and will work for online and offline migartions.
> > https://review.openstack.org/87935 But I faced the problem that Metering
> > usage won't be fixed as we need to create 2 tables (meteringlabels,
> > meteringlabelrules). I tried to create both in patch set #7 but it won't
> > work for offline migration. In fact to fix Grenade it is enough to
> > create meteringlabels table, that is done in my change in the last patch
> > set #8. I want to ask reviewers to take a look at this change and
> > suggest something or approve it. I'm available on IRC(akamyshnikova) or
> > by email.
> >
> > Regards
> > Ann
> Hi Ann,
>
> Good suggestion how to get out of failing job but I don't think it
> should go to 33c3db036fe4_set_length_of_description_field_metering.py
> because this failure is grenade specific while the real issue is a fact
> that we're not able to add new service plugin to already deployed Neutron.
>
> I think the same workaround you proposed in the 87935 review should go
> to grenade itself (from-havana/upgrade-neutron script) just to have the
> job working on havana->icehouse upgrade. It's a bit of ugly workaround
> though but imho so far the best solution to reach stable job in a short
> time.
>
> Kuba
>
> >
> >
> > _______________________________________________
> > OpenStack-dev mailing list
> > OpenStack-dev at lists.openstack.org
> > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
> >
>
>
> _______________________________________________
> OpenStack-dev mailing list
> OpenStack-dev at lists.openstack.org
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20140422/f44d500c/attachment.html>


More information about the OpenStack-dev mailing list