Re: [OSR-devel] GlusterFS Support

[Marc G. <gr...@at...>]

On Saturday 22 November 2008 22:37:00 Gordan Bobic wrote:
> Marc Grimme wrote:
> > On Saturday 22 November 2008 16:51:31 Gordan Bobic wrote:
> >> Marc Grimme wrote:
> >>> But why not go this way:
> >>> There is a method/function "clusterfs_services_start" which itself
> >>> calls ${clutype}_services_start which should be implemented by
> >>> etc/${rootfs}-lib.sh. Your $rootfs should be glusterfs and therefore
> >>> there should be a library glusterfs-lib.sh. The function
> >>> clusterfs_services_start should do all the preparations to mount the
> >>> clusterfilesystem and could also mount a local filesystem as a
> >>> prerequisite. So I would put it there. A rootfs can be specified as
> >>> follows:
> >>> <rootvolume name="/dev/whicheverdeviceyoulike" fstype="glusterfs"/>
> >>
> >> OK, this solution wins on the grounds that remapping the chrootenv seems
> >> to confuse glusterfs.
> >>
> >> But ${clutype}_services_start is dependent on ${clutype}, which is
> >> returned from getCluType(), which always returns "gfs". Will that not
> >> break things?
> >
> > Not any more take the latest version of clusterfs-lib.sh and it is
> > dependent on rootfs. I changed it how it should recently.
>
> Do you mean you changed getCluType(), or made it irrelevant? I freshly
> rebuilt the test box this morning, that that's where getCluType() was
> always returning "gfs". Anyway, I'll just ignore it if it's not important.
I wouldn't say that it's not important (although there are no other options 
implemented yet) but anyway it is "thought" as follows:
clutype should return the "clustertype" we are talking of. When using a 
cluster with /etc/cluster/cluster.conf it is a "gfs" cluster. Although I 
don't like the "gfs". But I cannot change it so easily so we have to stay 
with it for sometime. In next versions we are planing to change it how it 
should be to something like "redhat-cluster". But the "gfs" has to stay for 
the first as clustertype for any /etc/cluster/cluster.conf based cluster.

"rootfs" is a little bit ahead as we are supporting more then one rootfs (gfs, 
ocfs2 and ext3 .. gluserfs ;-) ). As we implemented ocfs2 and ext3 recently 
we had to review the use of "clutype" and "rootfs" as it should be.

So in your case clutype will stay "gfs" (yes I don't like it either but we'll 
change it some versions ahead. At latest when we are pushing sles10 in 
productive state. Until now you have to use /etc/cluster/cluster.conf in 
sles10 although there is no gfs or anything the like.) and "rootfs" should 
be "glusterfs". Therefore you have to build a glusterfs-lib.sh with all 
functions  called by clusterfs-lib.sh implemented. We will also need a 
listfile for it. But you've done this before so I don't expect it to be of 
any problem.

If you like I'll send you a list of functions to implement tomorrow. Ok?

-- 
Gruss / Regards,

Marc Grimme
http://www.atix.de/               http://www.open-sharedroot.org/