[Pacemaker] Do I need a resource for a clustered LVM volume group or not?

[Mike Diehn]

On Thu, Dec 9, 2010 at 4:47 AM, Andrew Beekhof <andrew at beekhof.net> wrote:

> On Thu, Dec 9, 2010 at 9:04 AM, Florian Haas <florian.haas at linbit.com>
> wrote:
> > On 2010-12-08 23:17, Mike Diehn wrote:
> >> While I waited to hear back from you, I tried this on our test cluster:
> >>
> >>   # bunch of primitives, then...
> >>   group Shared-FS dlm clvm o2cb vg1 fs1
> >>   clone Shared-FS_clone meta interleave="true"
> >>
> >> and it seems to work.
> >>
> >> Do you suppose I'm asking for trouble having all five resources in that
> >> one group?
> >
> > Now I know that cloning groups is supposed to work (it requires that all
> > the primitives within the group are themselves cloneable, but you've got
> > that covered), but I've never used it myself and can't say for sure that
> > it's working as advertised. But it's supposed to, and if it is not,
> > please let Andrew know.
>
> lmb uses it all the time and hasn't been complaining for quite a while.
> i'd claim the status is "known to work"
>

I've seen the group clone work in other clusters I have running, so I would
also.

I was wondering about practical concerns more than technical ones.  Like
"don't do that because you don't be able to unmount the filesystem without
stopping the whole" and such.

Also, Florian's point about DAV locking is well taken here.  I'll limit the
Apache to one node and use the other for the many, many svnadmin, svndump,
svnload tasks the repo admins run.

Thanks again,
Mike

--
Mike Diehn
Senior Systems Administrator
ANSYS, Inc - Lebanon, NH Office
mike.diehn at ansys.com, (603) 727-5492
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.clusterlabs.org/pipermail/pacemaker/attachments/20101209/150b6ac2/attachment-0002.htm>