LeftHand Multi-Site SAN

[Capra]

Is it possible to create Multi-Site SAN using two HP LeftHand P4300 4.8TB SAS Starter SAN Solution? Can someone explain me what is going on after primary site fails? Failover manager exist on logical 3rd site. What is procedure for activation my backup servers on second site?

Thanks!

[René Loser · Accepted solution]

Hello,
Yes, you can create a Multi-Site Cluster with 2 Nodes. Each can be in a different subnet.
I case of a node- or site-failure the FOM add the quorum to the remaining node. The volumes (if Network RAID2)will stay online. The iSCSI Initiators will connect to the VIP that will provided from the remaining node.
No manual intervention is necessary.

regards,
reNe
HP PreSales Switzerland

[Capra]

Thanks!
It was useful.
One more thing, when i have primary site failure, is it enough to just start my servers on secondary site and they will see lefthand over VIP. Of course i have before presented those same luns to my backup servers and powered off them.

[René Loser]

If your servers are not clustered that's correct. If you have a cluster the failover will be automatically.

regards,
reNe

[Capra]

Thank, i have no further questions. :)

[teledata]

Do be aware, if you intend to use this for an "always on" server or VM that the single VIP can take several seconds to "relocate" to a new storage node.

If you are running latency sensitive servers like Exchange/SQL or other database servers that that amount of time will likely result in a failed service.

If this is the case you should create a true multi-site SAN which consists of 2 VIPs (they have to each be in a unique subnet).

That way you can give your servers 2 redundant storage paths (one to each VIP) and storage path failover is very fast (usually very nearly instant), where the VIP loss, failure, and relocation can take several seconds.

If you are manually restarting the servers at the 2nd site, or using file/print servers (which are fairly tolerant of latency) then you don't need the dual-VIP MultiSite configuration

[someone847]

Hi Teledata, you mention a short timeout - have you witnessed such a short timeout (5 seconds or less) after a node failire (or simulated by removal of node from LAN).

We are running a 4 node multi-site, all RAID10, tested with single VIP and with 2 VIPs, we always get 30 second I/O outage.

Cheers.

[teledata]

That doesn't sound right.

I know we were getting less than a 30 second timeout when we did our testing.

The customer was having problems with Exchange (info store) crashing when they had a site link go down, and we converted them from a single VIP to a dual VIP multi-site configuration and did several tests, and our VMs, databases, and Exchange info store stayed up when we simulated Node or link failure.

Are you using this with VMware? If so can you verify that you see storage path redundancy for the datastores?

[someone847]

Hi Teledata, no VMware for us, all physical. We can see the DSM connections from each server to every Lefthand node. No combination of single/multi site, single/dual VIP, MPIO/not impacts our issue at all. Q. How were you measuring your I/O outage times? We're using IOmeter v2008.06.18 with read/writes and viewing the results display meter for total I/Os per second, since start of test, update every 1 second.

Cheers.