Hi,

As part of the Server lucid seeds blueprint [1] I'd like to request your
feedback on whether redhat-cluster-suite should be demoted to universe.

Pros for universe demotion:
* extremely difficult to test
* earlier versions were very painful to get security updates applied

Pros for keeping in main:
* package is actively maintained in Debian
* there is certainly user demand for it

[1]: https://blueprints.edge.launchpad.net/ubuntu/+spec/server-lucid-seeds

--
Mathias Gug
Ubuntu Developer http://www.ubuntu.com
--
ubuntu-server mailing list
ubuntu-server@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-server
More info: https://wiki.ubuntu.com/ServerTeam

"Mathias Gug" <mathiaz@ubuntu.com> wrote:

>Pros for universe demotion:
> * extremely difficult to test
> * earlier versions were very painful to get security updates applied

I think these are better arguments for removal than demotion. If it's too hard for Canonical to maintain, punting it to the community isn't likely to end well.

Scott K
--
ubuntu-server mailing list
ubuntu-server@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-server
More info: https://wiki.ubuntu.com/ServerTeam

On Mon, Jan 4, 2010 at 12:16 PM, Mathias Gug <mathiaz@ubuntu.com> wrote:
> Hi,
>
> As part of the Server lucid seeds blueprint [1] I'd like to request your
> feedback on whether redhat-cluster-suite should be demoted to universe.
>
> Pros for universe demotion:
> ** extremely difficult to test
> ** earlier versions were very painful to get security updates applied
>

I agree here, the testing isn't convenient and if that could be worked
out we can keep it, otherwise I suggest demotion as well

> Pros for keeping in main:
> ** package is actively maintained in Debian
> ** there is certainly user demand for it
>
> [1]: https://blueprints.edge.launchpad.net/ubuntu/+spec/server-lucid-seeds
>
> --
> Mathias Gug
> Ubuntu Developer *http://www.ubuntu.com
>
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.9 (GNU/Linux)
>
> iEYEARECAAYFAktCInwACgkQM0thG+z3pVgMPgCfZUobUw5yIS yL6b6meDqKkWs7
> 1ZUAoL+nA8kPttfFg3/3Kxu9AvYuTAuS
> =N1nz
> -----END PGP SIGNATURE-----
>
> --
> ubuntu-server mailing list
> ubuntu-server@lists.ubuntu.com
> https://lists.ubuntu.com/mailman/listinfo/ubuntu-server
> More info: https://wiki.ubuntu.com/ServerTeam
>

--
Regards,
Vikram Dhillon

~~~
There are lots of Linux users who don't care how the kernel works, but
only want to use it. That is a tribute to how good Linux is.
-- Linus Torvalds

--
ubuntu-server mailing list
ubuntu-server@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-server
More info: https://wiki.ubuntu.com/ServerTeam

On Mon, Jan 04, 2010 at 06:59:04PM -0500, Scott Kitterman wrote:
> "Mathias Gug" <mathiaz@ubuntu.com> wrote:
>
> >Pros for universe demotion:
> > * extremely difficult to test
> > * earlier versions were very painful to get security updates applied
>
> I think these are better arguments for removal than demotion. If it's too
> hard for Canonical to maintain, punting it to the community isn't likely
> to end well.

I don't think there was any implication in Mathias' email that this is about
Canonical or the community, only main vs. universe.

--
- mdz

--
ubuntu-server mailing list
ubuntu-server@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-server
More info: https://wiki.ubuntu.com/ServerTeam

On 04.01.2010 18:16, Mathias Gug wrote:

> As part of the Server lucid seeds blueprint [1] I'd like to request your
> feedback on whether redhat-cluster-suite should be demoted to universe.

Do *NOT* demote it to universe. There are libraries (libdlm,
libdlmcontrold) and maybe some binaries (gf2-tools) that we need in main
for new cluster stack.

Some binaries built from redhat-cluster source should be demoted to
universe, once new stack is promoted to main.

--
ubuntu-server mailing list
ubuntu-server@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-server
More info: https://wiki.ubuntu.com/ServerTeam

Hi Ante,

2010/2/3 Ante Karamatić <ivoks@grad.hr>:
> On 04.01.2010 18:16, Mathias Gug wrote:
>
>> As part of the Server lucid seeds blueprint [1] I'd like to request your
>> feedback on whether redhat-cluster-suite should be demoted to universe.
>
> Do *NOT* demote it to universe. There are libraries (libdlm,
> libdlmcontrold) and maybe some binaries (gf2-tools) that we need in main
> for new cluster stack.
>

I had a discussion with Fabio, the upstream redhat-cluster-suite
maintainer, and it seems that the stack corosync+openais+pacemaker is
not feature equivalent with the current redhat-cluster-suite. Moreover
redhat-cluster-suite is currently updated to use the new components
but is not ready yet.
It seems that the current redhat-cluster-suite is phased out and
maintenance from upstream will stop which is not the best option for
an LTS release. And the new cluster stack is being worked on but
hasn't stabilized yet. So right now it seems we're stuck between two
options. I wonder if demoting redhat-cluster-suite for lucid while
keeping the new stack in universe would be the best option for the
time frame that covers Lucid (LTS)?

Once the new stack has stabilized (and redhat-cluster-suite has been
updated to use the new components) we should reconsider promoting it
to main.


--
Mathias Gug
Ubuntu Developer http://www.ubuntu.com

--
ubuntu-server mailing list
ubuntu-server@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-server
More info: https://wiki.ubuntu.com/ServerTeam

On 07.02.2010 19:01, Mathias Gug wrote:

> I had a discussion with Fabio, the upstream redhat-cluster-suite
> maintainer, and it seems that the stack corosync+openais+pacemaker is
> not feature equivalent with the current redhat-cluster-suite. Moreover
> redhat-cluster-suite is currently updated to use the new components
> but is not ready yet.

Well, parts of cluster suite will be unmaintained. I doubt support for
distributed lock manager or gfs2 will be dropped (and those are only
things we need). Stuff that will go away are cman, rgmanager, etc...

> options. I wonder if demoting redhat-cluster-suite for lucid while
> keeping the new stack in universe would be the best option for the
> time frame that covers Lucid (LTS)?

I would really like to see new stack in main. I've spent last two-three
weeks testing it, adopting other packages and I'd rather use new stack
than any version of cman (RHCS).

To be clear, I do have strong feelings about this.

--
ubuntu-server mailing list
ubuntu-server@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-server
More info: https://wiki.ubuntu.com/ServerTeam

Hi Ante,

2010/2/7 Ante Karamatić <ivoks@grad.hr>:
>
> Well, parts of cluster suite will be unmaintained. I doubt support for
> distributed lock manager or gfs2 will be dropped (and those are only
> things we need). Stuff that will go away are cman, rgmanager, etc...
>

Agreed - it's my understanding as well. DLM and GFS2 will be part of
the new cluster stack and thus be maintained. While cman and rgmanager
should be replaced with pacemaker and some other daemons.

If I understand correctly it's currently impossible to upgrade
automatically from cman/rgmanager to the new stack. Fabio told me that
there were plans to add support for the existing cman/rgman
configuration syntax to the new components (pacemaker) but that
feature was not ready yet.

So how would one upgrade from Hardy rhcs to the new cluster stack in Lucid?

> I would really like to see new stack in main. I've spent last two-three
> weeks testing it, adopting other packages and I'd rather use new stack
> than any version of cman (RHCS).
>
> To be clear, I do have strong feelings about this.

I understand your point of view and appreciate the effort you've put
into testing the new stack - which is already in universe IIUC. I'm
trying to figure out what we'd loose with the new cluster stack and
how upgrades from Hardy rhcs should be handled.

I hope that through out the discussion we'll be able to reach a
consensus on which cluster stack options makes the most sense for the
next LTS release.

--
Mathias Gug
Ubuntu Developer http://www.ubuntu.com

--
ubuntu-server mailing list
ubuntu-server@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-server
More info: https://wiki.ubuntu.com/ServerTeam

On 07.02.2010 20:52, Mathias Gug wrote:

> If I understand correctly it's currently impossible to upgrade
> automatically from cman/rgmanager to the new stack. Fabio told me that
> there were plans to add support for the existing cman/rgman
> configuration syntax to the new components (pacemaker) but that
> feature was not ready yet.

It's also impossible to do heartbeat -> pacemaker upgrade without any
user interaction.

> So how would one upgrade from Hardy rhcs to the new cluster stack in Lucid?

One would get newer rhcs, which would be in universe (well, parts of it
anyway).

Having automatic configuration merge would be fantastic, but I don't see
it as a requirement. We are talking about two different products here.
Did we (desktop) merge pidgin's config when it was replaced? We should
provide good documentation for that and some of it can be done by simple
editing of test cases which I wrote
(https://wiki.ubuntu.com/ClusterStack/LucidTesting).

> I understand your point of view and appreciate the effort you've put
> into testing the new stack - which is already in universe IIUC. I'm
> trying to figure out what we'd loose with the new cluster stack and
> how upgrades from Hardy rhcs should be handled.

No, stuff that's in universe is very old and stack changed a lot since
then. If it stays in universe, we'll lose support for pacemaker both in
OCFS2 and GFS2. It's main or nothing. Having both crippled pacemaker
cluster stack and redhat-cluster (which will for sure reach its EOL
before 10.04 does) would make Ubuntu quite undesirable for fail over
systems.

--
ubuntu-server mailing list
ubuntu-server@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-server
More info: https://wiki.ubuntu.com/ServerTeam

On 07.02.2010 19:01, Mathias Gug wrote:

> I had a discussion with Fabio, the upstream redhat-cluster-suite
> maintainer, and it seems that the stack corosync+openais+pacemaker is
> not feature equivalent with the current redhat-cluster-suite.

Btw, pacemaker supports GFS2 and with RHCS 3.0.7 one can create and
mount GFS2 file systems.

--
ubuntu-server mailing list
ubuntu-server@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-server
More info: https://wiki.ubuntu.com/ServerTeam