[PVE-User] GlusterFS VM failover

Leslie-Alexandre DENIS infoslad at gmail.com
Tue Jan 28 19:00:18 CET 2014


Le 28/01/2014 17:52, Angel Docampo a écrit :
> Hi there,
>
> Mmmm, if you use /etc/fstab or glusterfs client, you will be accessing 
> the gluster via FUSE.
>
> And that s*cks.
>
> I do a trick on my proxmox cluster, as well gluster cluster. I do have 
> one 10Gb interface dedicated to gluster on each node, and another 1Gb 
> interface dedicated to proxmox cluster. So, in hosts, my config is 
> more or less this one.
>
> #PROXMOX NODES
> 10.0.0.1 pve01
> 10.0.0.2 pve02
> #GLUSTER NODE
> 192.168.100.10  g01
> 192.168.100.20  g02
>
> Until this point, completely normal, now, on the first node 
> (pve01/g01) y put on hosts
> 192.168.100.10 gluster
>
> And this other hosts line on node pve02/g02
> 192.168.100.20 gluster
>
> So, on proxmox GUI I mount gluster:VOLUMENAME. Each node mounts its 
> own mountpoint and on the redhat cluster, the resource is called the 
> same: gluster.
>
> Its a trick if your virtualization nodes are also storage nodes, which 
> is my case.
> Hope it helps.
>
Thanks for the input Angelo, it's a good lightweight tip but 
unfortunately it doesn't fit my needs. You can play with CARP (uCARP) to 
play with VIP too.

By default GlusterFS does the automatic failover between nodes but if 
your primary node crashes completely, regarding to networking or OS 
layer, and your client reboots it will be completely stuck. The fact to 
specify a backup address explicits the second node's and prevents that.

Regards
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.proxmox.com/pipermail/pve-user/attachments/20140128/45d64327/attachment.htm>


More information about the pve-user mailing list