|
Home > Archive > Debian Developers > July 2004 > Packaging GFS/Redhat Cluster
You are viewing an archived Text-only version of the thread.
To view this thread in it's original format and/or if you want to reply to
this thread please [click here]
| Author |
Packaging GFS/Redhat Cluster
|
|
| Joerg Wendland 2004-07-11, 5:51 pm |
| Hi *,
I'm currently trying to package GFS and the associated clustering
tools[0]. Bastian Blank already files #257249 but does not respond to
it. I would like to create an alioth project (I'd do that at once but I
don't want to steal Bastian's ITP without consens) and have a team
maintain these packages especially as I have no experience in building
kernel-patch packages. Likewise I am (or the team will be) in need of
coordination with lvm2 and devmapper maintainer (Patrick Caulfield,
CC'ed for that) because the cluster framework brings patches for those.
So what I want now before I begin:
- Bastian, what is your status regarding your ITP?
- Patrick, how can we work together?
- Anybody wanting to help? Especially with the kernel-patches?
What I will provide for development is a hardware testbed to eliminate
the need for obscure NDB or similar setups.
Thanks for your attention,
Joerg
[0] http://sources.redhat.com/cluster
--
Joerg "joergland" Wendland | http://www.wendlandnet.de/joerg/
GPG: 51CF8417 FP: 79C0 7671 AFC7 315E 657A F318 57A3 7FBD 51CF 8417
| |
| Bastian Blank 2004-07-11, 5:51 pm |
| On Sun, Jul 11, 2004 at 07:07:46PM +0200, Joerg Wendland wrote:
> - Bastian, what is your status regarding your ITP?
Packages are finished except of the descriptions and I wait for 64bit
libxml2/libz1 as I decided to only support sparc64 and s390x because
there is no biarch support and it does not work on s390.
The repository is http://lophos.multibuild.org/svn/cluster.
> - Anybody wanting to help? Especially with the kernel-patches?
I don't think someone wants to write biarch support for it.
> What I will provide for development is a hardware testbed to eliminate
> the need for obscure NDB or similar setups.
There exists a special nbd variant for use with gfs. It is currently not
packaged by that is easy.
Bastian
--
War is never imperative.
-- McCoy, "Balance of Terror", stardate 1709.2
| |
| Bastian Blank 2004-07-11, 5:51 pm |
| On Sun, Jul 11, 2004 at 07:07:46PM +0200, Joerg Wendland wrote:
> Bastian Blank already files #257249 but does not respond to
> it.
Hu?
> I would like to create an alioth project
You mean that machine called haydn with broken hardware?
> (I'd do that at once but I
> don't want to steal Bastian's ITP without consens) and have a team
> maintain these packages especially as I have no experience in building
> kernel-patch packages.
Patrick Caulfield accepted the invitation to the team.
> - Patrick, how can we work together?
Maybe it will be usefull to have lvm2/devmapper managed by the same
team.
Bastian
--
Leave bigotry in your quarters; there's no room for it on the bridge.
-- Kirk, "Balance of Terror", stardate 1709.2
| |
| Joerg Wendland 2004-07-11, 5:51 pm |
| Bastian Blank, on 2004-07-11, 19:32, you wrote:
> On Sun, Jul 11, 2004 at 07:07:46PM +0200, Joerg Wendland wrote:
>
> Packages are finished except of the descriptions and I wait for 64bit
> libxml2/libz1 as I decided to only support sparc64 and s390x because
> there is no biarch support and it does not work on s390.
>
> The repository is http://lophos.multibuild.org/svn/cluster.
Great, so I don't have to do anything :-)
>
> I don't think someone wants to write biarch support for it.
Is this an issue? I am not into this biarch thing but do these few libs
and tools have to biarch to be useful?
>
> There exists a special nbd variant for use with gfs. It is currently not
> packaged by that is easy.
Yes, GNBD, but it's not stable enough for production use and brings an
additional point of failure to your system. It may be useful for
testing GFS nevertheless.
Joerg
--
Joerg "joergland" Wendland | http://www.wendlandnet.de/joerg/
GPG: 51CF8417 FP: 79C0 7671 AFC7 315E 657A F318 57A3 7FBD 51CF 8417
| |
| Joerg Wendland 2004-07-11, 5:51 pm |
| Bastian Blank, on 2004-07-11, 19:37, you wrote:
> On Sun, Jul 11, 2004 at 07:07:46PM +0200, Joerg Wendland wrote:
>
> Hu?
I wrote only to the BTS.
>
> You mean that machine called haydn with broken hardware?
I do.
> Maybe it will be usefull to have lvm2/devmapper managed by the same
> team.
I think so.
Joerg
--
Joerg "joergland" Wendland | http://www.wendlandnet.de/joerg/
GPG: 51CF8417 FP: 79C0 7671 AFC7 315E 657A F318 57A3 7FBD 51CF 8417
| |
| Patrick Caulfield 2004-07-12, 2:50 am |
| On Sun, Jul 11, 2004 at 07:37:25PM +0200, Bastian Blank wrote:
> On Sun, Jul 11, 2004 at 07:07:46PM +0200, Joerg Wendland wrote:
>
> Hu?
>
>
> You mean that machine called haydn with broken hardware?
>
>
> Patrick Caulfield accepted the invitation to the team.
>
>
> Maybe it will be usefull to have lvm2/devmapper managed by the same
> team.
Certainly lvm is closely integrated in the clustering though I've organised it
so that the clustering for LVM can be packaged seperately to the main LVM
binaries - it's basically a daemon, a shared library and a config file change.
The bit I haven't yet decided is how to do the biarch bits. lvm2 is currently
built 32bit for sparc64 et al, but will need to be 64bit for it to be
clustered, which implies a 64bit device mapper library too. I suspect the best
thing to do is to built 32 and 64 bit libdevmapper and then move to building
lvm2 64bit only.
--
patrick
--
To UNSUBSCRIBE, email to debian-devel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
| |
| Joerg Wendland 2004-07-12, 7:53 am |
| Patrick Caulfield, on 2004-07-12, 08:21, you wrote:
> Certainly lvm is closely integrated in the clustering though I've organised it
> so that the clustering for LVM can be packaged seperately to the main LVM
> binaries - it's basically a daemon, a shared library and a config file change.
Yea. Having GFS without the clustered LVM is a non-option. Different
packages with and without cluster support do make sense with regard to
many people not needing the cluster infrastructure when using LVM.
Joerg
--
Joerg "joergland" Wendland | http://www.wendlandnet.de/joerg/
GPG: 51CF8417 FP: 79C0 7671 AFC7 315E 657A F318 57A3 7FBD 51CF 8417
| |
| Ondøej Surý 2004-07-12, 7:53 am |
| On Mon, 2004-07-12 at 14:20 +0200, Joerg Wendland wrote:
> Patrick Caulfield, on 2004-07-12, 08:21, you wrote:
>
> Yea. Having GFS without the clustered LVM is a non-option. Different
> packages with and without cluster support do make sense with regard to
> many people not needing the cluster infrastructure when using LVM.
Does clustered enhanced LVM change existing functionality of LVM (lvm2)?
If it can be *safely* added there's no reason to make separate package.
O.
--
Ondřej Surý <ondrej@sury.org>
| |
| Patrick Caulfield 2004-07-12, 5:58 pm |
| On Mon, Jul 12, 2004 at 02:57:56PM +0200, Ond?ej Surý wrote:
> On Mon, 2004-07-12 at 14:20 +0200, Joerg Wendland wrote:
>
> Does clustered enhanced LVM change existing functionality of LVM (lvm2)?
> If it can be *safely* added there's no reason to make separate package.
Not really. It adds another "locking" method to the lvm binary, that's all -
which will never be called in a single-host environment. Or it can be added as
a shared library. There's a daemon too that needs to be running if the nodes
are clustered.
One of the neat things about having it in a seperate package is that the
installation can modify the lvm config file to switch on the cluster add-ons,
rather than relying on the user doing it. Though I don't have any string
opinions one way or ther other.
--
patrick
--
To UNSUBSCRIBE, email to debian-devel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
|
|
|
|
|