On Fri, 09 May 2014 12:14:50 +0100
André Rodier <andre@rodier.me> wrote:
1) At the moment, I have only one computer available, so is it
possible to start a DRBD cluster with only one computer, and add a
second computer later.
Yes, but you can't test certain things. What you can do is to set up
Xen on your Machine and create two hosts for testing. The downside is
that if you doesn't want to keep Xen you have to set up anything again
when you second physical machine is available.
2) Is it really wise to use a so complex stack to host virtual
machines? 3) Would it be faster to use directly LVM volumes to host
virtual machines?
Works fine for me (except I don't use GFS2). As I use a
Failover-Cluster-Setup I have a classic primary/secondary-DRBD-Setup.
So there is no need for a Cluster-Filesystem, as only one DRBD-Drive
is active (mounted). I use ext4.
4) Would it be better to use another stack, for instance DRBD on top
of software RAID?
I'm using it that way (in fact I use a hardware RAID10, but did some
tests before using mdadm, works fine too.)
5) Is DRBD is the best solution to replicate LVM modules easily with
two computers?
I replicate the ext4, not the LVM. My Machines are Xen-DomUs having no
knowledge that their drives have LVM-Support at all. Only the Dom0s
have access to LVM. But from the top of my head I could not think of
any problems replicating the LVM itself.
My final goal is to be able to do live migration, between two or more
virtual machines hosts, without using a shared storage.
Using it for more than a year now, together with pacemaker/corosync,
without problems on about 35 machines.
Regards.