[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[kolla] State of ppc64le support

W dniu 16.09.2019 o 18:45, Michael Turek pisze:
> Hey all,
> We do use kolla. Let me see if I can shed some light here.

I have to admit that I wanted to check is there anyone using Kolla on
ppc64le. Thanks for replies.

>>         CentOS builds lack 'rabbitmq' 3.7.10 (we use external repo) and
>>         'gnocchi' binary images are not buildable due to lack of some
>>         packages
>>         (issue already reported to CentOS by TripleO team).
> We should be getting the gnocchi issue fixed. See this thread
> https://lists.centos.org/pipermail/centos-devel/2019-September/017721.html

I am in that thread ;)

> The rabbitmq issue is confusing me. The version provided for x86_64
> seems to be the same one provided for ppc64le, but maybe I'm missing
> something. If there's a package we need to get published, I can
> investigate.

External repo is used and we install "rabbitmq-3.7.10". One 'if ppc64le'
check and will work by using in-centos-repo version.

>>         Ubuntu builds lack MariaDB 10.3 because upstream repo is broken.
>>         Packages index is provided for 'ppc64le' but no packages so we
>>         get 404
>>         errors.
> Unfortunately I'm not well versed on the gaps in Ubuntu.

I am fine with it. No one noticed == no one uses.

> The question I have is, what do you need to maintain support? I can join
> this week's IRC meeting if that would be helpful.

For me a knowledge that someone is using is enough to keep it available.
Would not call it 'maintaining support' as I do builds on ppc64le once
per cycle (if at all per cycle).

> Also, last week mnasiadka reached out to me asking if we might be able
> to turn on kolla jobs in pkvmci (our third party CI -
> https://wiki.openstack.org/wiki/ThirdPartySystems/IBMPowerKVMCI ). I
> plan to talk to our CI folks this week to see if we have capacity for this.

Some kind of CI job would be great. Even simple 'centos/source' combo.

I have two patches adding AArch64 CI but we (Linaro) have to fix our
OpenStack cluster first. All Ceph nodes use hard drives only and
probably not configured optimally. As a result we are unable to fit in
three hours required by Zuul.