[one-users] Architecture advice

[Steven C Timm]

AT the moment we have an active/passive head node setup for OpenNebula with a SAN backend for the image repository. Opennebula service is managed by redhat clustering  as is the GFS2+CLVM file system.  We run a few virtual machines on the frontend machines as well.  If we were using an ssh-based transfer manager that would not work, nor did it work when we were using a gfs2+drbd file system for the image repo.

We are thinking instead to shift the ONE head node to be a live-migrateable virtual machine (or machines), the problem is how to get the VM to hook into the clustered file system with good bandwidth.

Steve Timm


From: users-bounces at lists.opennebula.org [mailto:users-bounces at lists.opennebula.org] On Behalf Of Shankhadeep Shome
Sent: Friday, March 28, 2014 4:53 PM
To: Stuart Longland
Cc: users
Subject: Re: [one-users] Architecture advice

What is your definition of large? This is a difficult question to answer without more details.

On Wed, Mar 26, 2014 at 7:08 PM, Stuart Longland <stuartl at vrt.com.au<mailto:stuartl at vrt.com.au>> wrote:
On 18/03/14 02:35, Gandalf Corvotempesta wrote:
> Hi to all
> i'm planning a brand new cloud infrastructure with opennebula.
> I'll have many KVM nodes and 3 "management nodes" where I would like
> to place OpenNebula, Sunstone and something else used to orchestrate
> the whole infrastructure
>
> Simple question: can I use these 3 nodes to power on OpenNebula (in
> HA-Configuration) and also host some virtual machines managed by
> OpenNebula ?
I could be wrong, I'm new to OpenNebula myself, but from what I've seen,
the management node isn't a particularly heavyweight process.

I had it running semi-successfully on an old server here (pre
virtualisation-technology).  I say semi-successfully; the machine had a
SCSI RAID card that took a dislike to Ubuntu 12.04, so the machine I had
as the master would die after 8 hours.

I was using SSH based transfers (so no shared storage) at the time.

Despite this, the VMs held up, they just couldn't be managed.  This
won't be the case if your VM hosts mount any space off the frontend
node: in which case a true HA set-up is needed.  (And lets face it, I
wouldn't recommend running the master node as a VM if you're going to be
mounting storage directly off it for the hosts.)

Based on this it would seem you could do a HA setup with some shared
storage between the nodes, either a common SAN or DR:BD to handle the
OpenNebula frontend.

Seeing as OpenNebula will want to control libvirt on the hosts (being
that you're also suggesting making these run the OpenNebula-managed VMs
too, this might have to be a KVM process managed outside of libvirt.
Not overly difficult, just fiddly.

But, as I say, I could be wrong, so take the above advice with a grain
of salt.

Regards,
--
Stuart Longland
Systems Engineer
     _ ___
\  /|_) |                           T: +61 7 3535 9619<tel:%2B61%207%203535%209619>
 \/ | \ |     38b Douglas Street    F: +61 7 3535 9699<tel:%2B61%207%203535%209699>
   SYSTEMS    Milton QLD 4064       http://www.vrt.com.au


_______________________________________________
Users mailing list
Users at lists.opennebula.org<mailto:Users at lists.opennebula.org>
http://lists.opennebula.org/listinfo.cgi/users-opennebula.org

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.opennebula.org/pipermail/users-opennebula.org/attachments/20140329/08a351e1/attachment-0002.htm>