[PVE-User] PVE 2.1 and HA with drbd...

Robert Fantini rob at fantinibakery.com
Fri Jul 27 17:13:00 CEST 2012

On 07/27/2012 08:55 AM, Gilberto Nunes wrote:
> Hi....
> Formerly, I always used heartbeat and  drbd to provide HA to my 
> Virtual Machines...
> I always configured heartbeat to check the physical nodes, and if one 
> of them crashing down, the heartbeat start qm to initialize the 
> formerly VM running on the node that crash...
> Now, I see that PVE bring to us corosync-pve...
> My question is: it's simple do the same work that heartbeat do before?
> Is there some risk if I change some configuration on the default 
> corosync-pve config???
> Thanks
> Cheers

We also used drbd + heartbeat in 1.9 .  And Debian etch before that..

Using a  Primary/Secondary drbd set up with something like heartbeat to 
control which system is the Primary just worked.

  It would be I think a very good feature to have in a 2-node drbd 
cluster.    Or is there something already built in to PVE cluster that 
deals with this?

  I understand the the full heartbeat package is not compatible with pve:

# aptitude install heartbeat
The following NEW packages will be installed:
   cluster-agents{a} cluster-glue{a} heartbeat libcluster-glue{a} 
libcorosync4{a} libesmtp5{a} libheartbeat2{a}
   libnet1{a} libopenhpi2{a} openhpid{a} pacemaker{a}
0 packages upgraded, 11 newly installed, 0 to remove and 2 not upgraded.
Need to get 2,968 kB of archives. After unpacking 10.4 MB will be used.
The following packages have unmet dependencies:
   libcorosync4-pve: Conflicts: libcorosync4 but 1.2.1-4 is to be installed.
The following actions will resolve these dependencies:

       Remove the following packages:
1)      clvm
2)      corosync-pve
3)      fence-agents-pve
4)      libcorosync4-pve
5)      libopenais3-pve
6)      libpve-access-control
7)      libpve-storage-perl
8)      openais-pve
9)      proxmox-ve-2.6.32
10)     pve-cluster
11)     pve-manager
12)     qemu-server
13)     redhat-cluster-pve
14)     resource-agents-pve
15)     vzctl
Accept this solution? [Y/n/q/?] q

  For now we are using   drbd Primary/Primary .
   We have the KVM's running on one node.

If the node we are using breaks, we'll do a manual switch over. To do so 
this was suggested by Dietmar on the forum:

1-First make sure the other node is really down!

2-Then set expected votes to gain quorum:  [ this may already be set in 
our 2 node cluster.conf ? ]
  # pvecm expected 1

3-Then move the config file to correct position:
# mv /etc/pve/nodes/<oldnode>/qemu-server/<vmid>.conf 

   The KVM disks are of course on both nodes thanks to drbd.   So only 
the .conf files need to me moved.

   I was never comfortable with heartbeat even after 5+ years for auto 
fail over.   So I do not mind doing the fail over manually in PVE 
version 2.

   PS:  in the future ,  sheepdog  seems like it will be a better way to 
ensure the survivability  of KVM disks .

> -- 
> Gilberto Nunes
> _______________________________________________
> pve-user mailing list
> pve-user at pve.proxmox.com
> http://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-user
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.proxmox.com/pipermail/pve-user/attachments/20120727/1b51218a/attachment.htm>

More information about the pve-user mailing list