11:08 a.m.
Hi,
I've been trying to use centos-release-openstack-mitaka repo to deploy undercloud,
build overcloud images (with some hacks to use the same repo) and deploy the overcloud.
The rationale behind it is to use Centos-provided repos with production deployment in
mind.
rdo-trunk-mitaka-tested repo has many bugs fixed, but I do not think it can be qualified
as production-ready repository.
Long story short, centos-release-openstack-mitaka has os-net-config package at version
0.2.2-1.el7, which broke Linux Bond configurations [1].
The issue had been fixed back in April and made it to RHEL OSP (Liberty), but not to
centos-release-openstack-mitaka.
I'm not sure if rdo-list is the best place to ask this question, but would appreciate
if you can point me towards resources describing how Centos stable repos for OpenStack are
maintained, tested and packages promoted to newer versions.
It feels as if today, one has a choice between a bit outdated Centos repo and too hot
rdo-trunk-tested. Is there anything in between that can be used for production
deployments?
Thanks,
Ilja
1. https://bugzilla.redhat.com/show_bug.cgi?id=1323717
2:59 p.m.
Hi Ilja,
On Mon, 2016-08-15 at 17:08 +0100, Ilja Maslov wrote:
Hi,
I’ve been trying to use centos-release-openstack-mitaka repo to
deploy undercloud, build overcloud images (with some hacks to use the
same repo) and deploy the overcloud.
I think you are running into the same issues we are.
The rationale behind it is to use Centos-provided repos with
production deployment in mind.
rdo-trunk-mitaka-tested repo has many bugs fixed, but I do not think
it can be qualified as production-ready repository.
Agreed, production deployment at the moment with RDO is not a pretty
picture. Sorry to be blunt but this is my experience after around a
year of working with it. We have to pin package versions, mash
repositories together, run deployments from nightly images and the
like. We generally resort to the OSP documentation for a definitive
source.
Long story short, centos-release-openstack-mitaka has os-net-config
package at version 0.2.2-1.el7, which broke Linux Bond configurations
[1].
We also have issues with incorrect galera affecting HA and a bug with
keystone v3 domain support
The issue had been fixed back in April and made it to RHEL OSP
(Liberty), but not to centos-release-openstack-mitaka.
I’m not sure if rdo-list is the best place to ask this question, but
would appreciate if you can point me towards resources describing how
Centos stable repos for OpenStack are maintained, tested and packages
promoted to newer versions.
I would also really appreciate this so there can be a cohesive effort
towards maintaining a stable rdo release.
It feels as if today, one has a choice between a bit outdated Centos
repo and too hot rdo-trunk-tested. Is there anything in between that
can be used for production deployments?
I think the idea with the CentOS repo is that it only gets updated with
when upstream release an update. But there doesn't seem to be any CI
that tests whether this setup actually works.
I have added this as a topic for Wednesday's IRC meeting:
https://etherpad.openstack.org/p/RDO-Meeting
but it is unlikely I will be able to attend this as I will be
travelling.
Thanks,
Ilja
1. https://bugzilla.redhat.com/show_bug.cgi?id=1323717
--
Regards,
Christopher
1:35 a.m.
On 15/08/16 21:59, Christopher Brown wrote:
Hi Ilja,
On Mon, 2016-08-15 at 17:08 +0100, Ilja Maslov wrote:
> Hi,
>
I think the idea with the CentOS repo is that it only gets updated with
when upstream release an update. But there doesn't seem to be any CI
that tests whether this setup actually works.
I have added this as a topic for Wednesday's IRC meeting:
https://etherpad.openstack.org/p/RDO-Meeting
but it is unlikely I will be able to attend this as I will be
travelling.
Thank you for your feedback, it really helps. If you have the feeling,
nobody listening, that's wrong.
Please share your experience, where you had to pin versions etc. It
would be good to avoid others running into the same issues and to
stabilize quicker.
About CI, other folks might want to comment on the state.
Matthias
--
Matthias Runge <mrunge(a)redhat.com>
Red Hat GmbH, http://www.de.redhat.com/, Registered seat: Grasbrunn,
Commercial register: Amtsgericht Muenchen, HRB 153243,
Managing Directors: Charles Cachera, Michael Cunningham,
Michael O'Neill, Eric Shander
2:58 a.m.
Hi Matthias,
On Tue, 2016-08-16 at 07:35 +0100, Matthias Runge wrote:
On 15/08/16 21:59, Christopher Brown wrote:
>
> Hi Ilja,
>
> On Mon, 2016-08-15 at 17:08 +0100, Ilja Maslov wrote:
> >
> > Hi,
> >
>
> I think the idea with the CentOS repo is that it only gets updated
> with
> when upstream release an update. But there doesn't seem to be any
> CI
> that tests whether this setup actually works.
>
> I have added this as a topic for Wednesday's IRC meeting:
>
> https://etherpad.openstack.org/p/RDO-Meeting
>
> but it is unlikely I will be able to attend this as I will be
> travelling.
>
Thank you for your feedback, it really helps. If you have the
feeling,
nobody listening, that's wrong.
Thanks, much appreciated. I am very keen to be part of a fix for this.
We do our fair share of OSP and RDO deployments and both have their
quirks but the ability to re-use config across both means that the
RDO/OSP model works very well for us.
Please share your experience, where you had to pin versions etc. It
would be good to avoid others running into the same issues and to
stabilize quicker.
Last time I think trown made fairly obvious suggestion that a bugzilla
at least would be good, therefore:
https://bugzilla.redhat.com/show_bug.cgi?id=1365884
to which ggillies compiled a good response here:
https://www.redhat.com/archives/rdo-list/2016-August/msg00127.html
but I have yet to see any feedback (I think it was a holiday in the US yesterday?)
This is where we have pinned galera to the earlier working version Graeme identified.
The keystone v3 error we have not logged as we have reverted
temporarily to v2 to carry out tempest tests. I will file a bug for
that once we replicate this again.
About CI, other folks might want to comment on the state.
I think really the issue is with building the images here. The stable
"openstack overcloud image build" doesn't appear to get tested as this
is broken. When I followed the "Install from Mitaka branch"
instructions here:
http://tripleo.org/basic_deployment/basic_deployment_cli.html#get-image
s
it pulled in packages from newton-testing, though that could have been
an error on my part.
I think what would be good would be to know who to contact, file bugs,
submit patches etc for the stable branch.
Matthias
--
Matthias Runge <mrunge(a)redhat.com>
Red Hat GmbH, http://www.de.redhat.com/, Registered seat: Grasbrunn,
Commercial register: Amtsgericht Muenchen, HRB 153243,
Managing Directors: Charles Cachera, Michael Cunningham,
Michael O'Neill, Eric Shander
_______________________________________________
rdo-list mailing list
rdo-list(a)redhat.com
https://www.redhat.com/mailman/listinfo/rdo-list
To unsubscribe: rdo-list-unsubscribe(a)redhat.com
--
Regards,
Christopher Brown
9:26 a.m.
So, how exactly newer packages get into CentOS repo? I see about 7 newer versions of the
os-net-config in various repos, ranging between versions 0.2.2 and 5.0.0. RHEL OSP errata
updates to 0.2.3, I believe.
Is there a place I can check to find out when os-net-config package will get updated in
CentOS repo? Do CentOS repos follow RHEL updates and errata?
Image creation is a very important subject, but if there is no stable repo, no matter what
process is used to create images, they will be as good as the repo the packages are pulled
from :)
Thanks,
Ilja
-----Original Message-----
From: rdo-list-bounces(a)redhat.com [mailto:rdo-list-bounces@redhat.com] On Behalf Of
Christopher Brown
Sent: Tuesday, August 16, 2016 3:59 AM
To: mrunge(a)redhat.com; rdo-list(a)redhat.com
Subject: Re: [rdo-list] centos-release-openstack-mitaka
Hi Matthias,
On Tue, 2016-08-16 at 07:35 +0100, Matthias Runge wrote:
On 15/08/16 21:59, Christopher Brown wrote:
>
> Hi Ilja,
>
> On Mon, 2016-08-15 at 17:08 +0100, Ilja Maslov wrote:
> >
> > Hi,
> >
>
> I think the idea with the CentOS repo is that it only gets updated
> with when upstream release an update. But there doesn't seem to be
> any CI that tests whether this setup actually works.
>
> I have added this as a topic for Wednesday's IRC meeting:
>
> https://etherpad.openstack.org/p/RDO-Meeting
>
> but it is unlikely I will be able to attend this as I will be
> travelling.
>
Thank you for your feedback, it really helps. If you have the feeling,
nobody listening, that's wrong.
Thanks, much appreciated. I am very keen to be part of a fix for this.
We do our fair share of OSP and RDO deployments and both have their quirks but the ability
to re-use config across both means that the RDO/OSP model works very well for us.
Please share your experience, where you had to pin versions etc. It
would be good to avoid others running into the same issues and to
stabilize quicker.
Last time I think trown made fairly obvious suggestion that a bugzilla at least would be
good, therefore:
https://bugzilla.redhat.com/show_bug.cgi?id=1365884
to which ggillies compiled a good response here:
https://www.redhat.com/archives/rdo-list/2016-August/msg00127.html
but I have yet to see any feedback (I think it was a holiday in the US yesterday?)
This is where we have pinned galera to the earlier working version Graeme identified.
The keystone v3 error we have not logged as we have reverted temporarily to v2 to carry
out tempest tests. I will file a bug for that once we replicate this again.
About CI, other folks might want to comment on the state.
I think really the issue is with building the images here. The stable "openstack
overcloud image build" doesn't appear to get tested as this is broken. When I
followed the "Install from Mitaka branch"
instructions here:
http://tripleo.org/basic_deployment/basic_deployment_cli.html#get-image
s
it pulled in packages from newton-testing, though that could have been an error on my
part.
I think what would be good would be to know who to contact, file bugs, submit patches etc
for the stable branch.
Matthias
--
Matthias Runge <mrunge(a)redhat.com>
Red Hat GmbH, http://www.de.redhat.com/, Registered seat: Grasbrunn,
Commercial register: Amtsgericht Muenchen, HRB 153243, Managing
Directors: Charles Cachera, Michael Cunningham,
Michael O'Neill, Eric Shander
_______________________________________________
rdo-list mailing list
rdo-list(a)redhat.com
https://www.redhat.com/mailman/listinfo/rdo-list
To unsubscribe: rdo-list-unsubscribe(a)redhat.com
--
Regards,
Christopher Brown
_______________________________________________
rdo-list mailing list
rdo-list(a)redhat.com
https://www.redhat.com/mailman/listinfo/rdo-list
To unsubscribe: rdo-list-unsubscribe(a)redhat.com
2834
days inactive
2835
days old
4 comments
3 participants
participants (3)
-
Christopher Brown -
Ilja Maslov -
Matthias Runge