[Debian-ha-maintainers] Cluster LVM daemon in Squeeze

Tim Small tim at seoss.co.uk
Wed Oct 13 19:46:44 UTC 2010


Hi,

Apologies if this is the wrong place to ask, feel free to tell me where 
to go....

I realise that clvm is not maintained by debian-ha-maintainers, but 
thought I'd try to sound out about it here.

One of the Redhat-supported configurations for multi-node with migration 
support for libvirt+kvm is to use clvm (multi-master DRBD or whatever 
kind of shared block device you'd like) - 
http://sources.redhat.com/cluster/wiki/KvmMigration

As built currently for Debian, clvm2 is built to use redhat cluster 
manager, it also seems to be quite poorly documented.  As cman takes 
over control of corosync, I'm guessing that it'll be difficult or 
impossible to use pacemaker on the same node (maybe you can use another 
corosync instance, or something, but I didn't see this documented).

I note that clvm has compile-time configure options for different 
cluster-managers, and that it currently seems to built with 
--with-clvmd=cman in squeeze.  These are the options:

--with-clvmd=TYPE       Build cluster LVM Daemon.
                           The following cluster manager combinations are valid:
                            * cman,gulm             (RHEL4 or equivalent)
                            * cman                  (RHEL5 or equivalent)
                            * cman,corosync,openais (or selection of them)
                            * singlenode            (localhost only)
                            * all                   (autodetect)
                            * none                  (disable build)

and I was wondering if one of the other options would be better in order 
to give more flexibility from an HA point-of-view?

I understand that SUSE also did some work to integrate clvmd with 
pacemaker, but I haven't looked into this, and I assume it'd be too-late 
for Squeeze, unless it's already been quietly integrated in the Squeeze 
version.

I understand that Redhat are planning on ditching RGManager etc. in 
favour of Pacemaker at some point, and we already run a couple of 
pacemaker-on-Debian clusters, so we'd probably prefer this way of 
running things too.  We could just drop clvmd entirely of course, but 
that'd lose a bit of ease + flexibility when it comes to managing 
storage from inside libvirt, so we'd like to stick with it if it's 
likely to be viable...

Cheers,

Tim.

-- 
South East Open Source Solutions Limited
Registered in England and Wales with company number 06134732.
Registered Office: 2 Powell Gardens, Redhill, Surrey, RH1 1TQ
VAT number: 900 6633 53  http://seoss.co.uk/ +44-(0)1273-808309




More information about the Debian-ha-maintainers mailing list