Gluster users - Jun 2016 - QEMU

On Mon, Jun 13, 2016 at 11:17:04PM +0200, Gandalf Corvotempesta
wrote:
> Hi to all
> anyone using gluster for qemu?
> I was looking at this:
>
https://raobharata.wordpress.com/2012/10/29/qemu-glusterfs-native-integration/
> 
> performance seems to be interesting, almost the same as local storage
> (pretty strange to me)
> 
> Anyone willing to share experience about this?
> all "vm" related platforms seems to looking at Ceph when shared
> storage is needed (and I don't understand why, gluster is much easier
> to build and maintain)
> _______________________________________________
> Gluster-users mailing list
> Gluster-users at gluster.org
> http://www.gluster.org/mailman/listinfo/gluster-users
Yep, using it with proxmox so it's qemu.
Performances seems good, but it does depend on the setup.
I know currently our production cluster is using 3.7.6 on 3 nodes
across 2 datacenters, with 10ms ping between them. Seemed mostly fine
at first but a client complained about performances, so I installed
a 3.7.11 cluster between 3 nodes across two datacenters closer together
(same town so not much ping between) and did some testing, it's a lot better
and the client was happy.
I don't know how much the update contributed, but I assume the ping is
playing a bit part in this.

I could send you the bonnie++ results from those two tests tomorrow if you want,
kept that at work.
I'll probably test that again with 3.7.12 just to compare. 

-- 
Kevin Lemonnier
PGP Fingerprint : 89A5 2283 04A0 E6E9 0111
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: Digital signature
URL:
<http://www.gluster.org/pipermail/gluster-users/attachments/20160614/47180763/attachment.sig>

2016-06-14 0:05 GMT+02:00 Kevin Lemonnier <lemonnierk at
ulrar.net>:
> Yep, using it with proxmox so it's qemu.
> Performances seems good, but it does depend on the setup.
> I know currently our production cluster is using 3.7.6 on 3 nodes
> across 2 datacenters, with 10ms ping between them. Seemed mostly fine
> at first but a client complained about performances, so I installed
> a 3.7.11 cluster between 3 nodes across two datacenters closer together
> (same town so not much ping between) and did some testing, it's a lot
better
> and the client was happy.
> I don't know how much the update contributed, but I assume the ping is
> playing a bit part in this.
Are you willing to share your hardware/network configuration? We are
planning a cluster
similiar to yours.
> I could send you the bonnie++ results from those two tests tomorrow if you
want,
> kept that at work.
Yes, this could be interested. If possible, could you please share
some benchmark for the
same VM running from gluster and running directly from localstorage ?

What i've read is that with native gluster connector for QEMU
performance are almost identical
to the plain qemu (this looks pretty strange to me)