[DRBD-user] DRBD and KVM for a HA-Cluster ?

Lentes, Bernd bernd.lentes at helmholtz-muenchen.de
Sat Jan 8 17:32:32 CET 2011

Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.


 
Digimer wrote:

> 
> On 01/07/2011 04:00 PM, Lentes, Bernd wrote:
> > VM1A and VM1B are identical VM's, offering the same service !
> 
> Here is the confusion. When you say "identical", you are 
> referring to the underlying data being mirrored on both 
> nodes, right? That is, one VM
> (VM1) that runs on *either* Node 1 *or* Node 2, right?

I mean both: the data is identical (that's the purpose of DRBD) and so, of course, the vm's are identical.

> 
> If you are thinking of two separate instances, however, that 
> is not possible.
> 
> 

"Otherwise, yes, that would work perfectly." I don't understand what you mean here.

> 
> > The vm's are installed in the plain DRBD, without any 
> filesystem. Each vm has just to be created once, because 
> DRBD, responsible for the replication, takes care that the vm 
> "appears"
> > on the other node. Right ? Is DRBD able to replicate that, 
> although there is no filesystem ?
> 
> Yes. DRBD is a block device level application. It doesn't 
> care at all about what might be using it. Think of DRBD as 
> just another hard drive.
> 
> > Is this scenario senseful ?
> 
> Yes.
> 
> > In this scenario, is it possible to run VM1A and VM1B 
> simultaneously, in a DRBD dual primary mode ? I don't think so.
> > I hope my scenario is clear.
> 
> Absolutely not. :)
> 
> VM1 is VM1. It's data will always be identical on both nodes, 
> but only one node will actually be hosting the VM at any one 
> time. Trying to start it up on the second Node would be a 
> *very* bad idea.
> 
> Remember;
> 
> Should you run into a split brain, you will need to manually 
> run 'drbdadm invalidate rX' on the node that the VM was /not/ 
> running on.
> Let's say that VM1 (DRBD r1) and VM3 (DRBD r3) were running 
> on Node 1 and VM2 (DRBD r2) was running on Node 2. Now the 
> DRBD resources split-brain (ie: network fault). You would now 
> need to run:
> 
> Node 1:
> - drbdadm invalidate r2
> 
> Node 2:
> - drbdadm invalidate r1
> - drbdadm invalidate r3
> 
> Once that is done, you could reconnect and the DRBD resources 
> would re-sync. Does this make sense?

Now things are becoming clearer. I try to avoid split brain by using two ethernet adapters on each node, combining them in a bond device.

Bernd

> 
> --
> Digimer
> E-Mail: digimer at alteeve.com
> AN!Whitepapers: http://alteeve.com
> Node Assassin:  http://nodeassassin.org
> 


More information about the drbd-user mailing list