[PVE-User] Proxmox on DRBD and OCFS2/GFS

Eneko Lacunza elacunza at binovo.es
Mon Aug 26 11:11:56 CEST 2013


Hi Marco,

On 26/08/13 10:53, Marco Gabriel - inett GmbH wrote:
> has anybody tried to use OCFS2 or GFS on top of DRBD for a Proxmox VE 
> setup?
>
> What I discovered so far:
>
> ·OCFS2 is not in the kernel, not built-in and not available as a module.
>
> ·GFS2 is in the kernel, but depends on cman package. Cman .deb package 
> conflicts with Proxmox .deb packages.
>
> Any ideas or experiences with such setups on Proxmox?
>
> From my point of view, this sounds like an ideal thing for a two node 
> setup.
>
>

We did test such a setup about 1-2 years ago. What we found was that:
- REALLY, you have to setup fencing right for cluster filesystems
- DRBD seemed to have troubles back then recovering properly after both 
nodes were shut down (ie, because a power outage, rack, rearrangement, 
whatever)

We finally discarded it (were trying active-active configuration) 
in-house, we felt it gave us more headaches than the peace-of-mind 
potentially gained by a shared-storage HA setup. We didn't have 
experience with DRBD so maybe it was us not setting up correctly.

Yet, we're using it in active-passive configuration in a client, though. 
They're quite happy, the only problem happened when both nodes went down 
after a cascade of power failures, we had to go onsite for manual 
intervention, but it was up and running in 30 minutes, with no data loss.

My advice is to carefully weight the need for a "shared" storage in your 
cluster against the added complexity. Also, is you're writing from both 
nodes at the same time, your I/O will suffer a lot.

Cheers
Eneko

-- 
Zuzendari Teknikoa / Director Técnico
Binovo IT Human Project, S.L.
Telf. 943575997
       943493611
Astigarraga bidea 2, planta 6 dcha., ofi. 3-2; 20180 Oiartzun (Gipuzkoa)
www.binovo.es

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.proxmox.com/pipermail/pve-user/attachments/20130826/e3318bdb/attachment.htm>


More information about the pve-user mailing list