[PVE-User] Planning an HA infrastructure

Eneko Lacunza elacunza at binovo.es
Fri Jan 24 10:44:56 CET 2014


Hi Philippe,

On 24/01/14 10:23, Philippe Schwarz wrote:
>> Hi Philippe,
>>
>> Maybe you have thought about this, but does it make sense for your
>> use case to have a PVE-HA cluster, that is dependent on a third
>> storage non-HA server?
>>
>> Do you REALLY need the HA capacity?
>>
>> I think that the 10 Gbit network connection will help most
>> connecting PVE servers with storage backend, be it another PVE
>> server or a storage one.
> Hi,
> thanks for your (clever) answer.
>
> I thought i could use the (one of the amazing) features of ZFS (like
> zfs send zfs receive) to avoid buying a costly second storage server.
> Costly because of the amount of disks.
>
> So having a 2nodes cluster with a master/slave storage SAN, was my idea.
>
> But, you're right: A HA cluster for pve, with a non-HA storage is a
> non-sense.
>
> But, i'm not confident with the idea of virtualizing the ZFS storage,
> loosing by the way the rock-solid, direct access to hardware features
> and self-healing capabilities of ZFS.
>
> Another question:In case of blade system; is the fencing devicing
> still useful ?
Never used a blade system myself, but I think that you need fencing 
anyway for a cluster (maybe the blade system provides a fencing system).
> Ok, let's assume i don't need the HA and will migrate manually a VM.
> I'm in a secondary school and can afford a few minutes (maybe hours if
> it occurs once a year maximum ..) downtime.
>
> What does it change to my infrastructure questions ?
Proxmox allows you to migrate disk from one storage to another with the 
VM up. You will notice slow-down in disk access, of course; but no 
downtime. Also you can migrate active VMs from one PVE server to another 
if using shared disk.

Otherwise, moving a full VM from local disk in a PVE server to another 
PVE server, will depend on the size of the disks, disk read and write 
performance, and network connection between PVE nodes. Using SATA 7K2 
disks and gigabit interconnect, I think that our ~100GB disk VM takes 
about 1h-1h30 to migrate from one PVE server to another.

Anyway, I'm not sure you're doing the right questions. Migration of VMs 
is an issue, but only if you need to move the VMs between PVE servers 
when they are working.

But HA usually is to protect yourself from hardware failure. If you 
don't have HA and a PVE server goes down, then you can't migrate the VM; 
you may create a new one with his disks, if disks are not in the PVE 
server; otherwise you must recover from backup, or fix the failed 
server, or put the failed server disks in another server if they are ok.

For me the critical part is the hard disk storage, so you must plan to 
have at least 2 hot copies on different servers to be protected from Hw 
failure. That will help recovering from failure, even if you have to do 
it by hand. I'm sorry to say that having a unique ZFS server won't help 
with that! If the server crashes you don't have access to the disks :)

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




More information about the pve-user mailing list