Advice on fully redundant Hyper-V cluster

Printable View

07-29-2011, 04:27 PM
Olias

Advice on fully redundant Hyper-V cluster

Hello to all you Open-e experts. I have been using DSS successfully in a Hyper-V environment. I currently have two separate Hyper-V servers and two DSS v6.00 up75 iSCSI SANs. Each 2008 R2 server is attached to one DSS iSCSI SAN via 10GB NICs and hosting several VMs.

I now would like to create a fully redundant setup using a Hyper-V cluster consisting of two clustered Hyper-V servers, two 10GB switches and two DSS iSCSI SANs. The final result would be a configuration in which ANY single device could fail but the overall system would continue to function.

Is this a possible/realistic expectation? Has anyone done this? I've seen the documentation for using DSS in a Windows Server cluster, but not using TWO DSS storage servers to appear as a single, redundant storage device for the cluster.
Any direction or advice would be greatly appreciated.
Thanks.
07-29-2011, 05:20 PM
To-M

By using the DSS V6 iSCSI Auto Failover to present one or more targets to the MS Cluster by using only 1 VIP "virtual IP Address" to connect to the DSS V6 Target. Setting up the iSCSI Failover links are below.

http://www.open-e.com/library/how-to-resources/
Open-E Replications

2011.03 Open-E DSS V6 Synchronous Volume Replication with Failover over a LAN (broadcast)

2011.03 Open-E DSS V6 Synchronous Volume Replication with Failover over a LAN (unicast)
07-30-2011, 01:42 AM
Olias

Great! I'll study up on those how-tos. Just to clarify: using the "Synchronous Volume Replication with Failover over a LAN" I can present the cluster with what appears as a single iSCSI storage unit that will pass all of the MS Cluster validations as a CSV (Cluster Shared Volume)?

Anyone out there actually doing this in the real world, or am I one of the first to "boldly go where no one has gone before"?
07-30-2011, 02:01 PM
enealDC

There are a number of people (myself included) who are already doing this. You aren't in uncharted waters. iSCSI failover and SCSI-3 Persistent reservations has been in Open-E for a while.
07-30-2011, 10:49 PM
jtucker

Broadcast vs unicast

I am also setting up failover with VMware - 2 open-e boxes (6 nic's each) and two vmware boxes (6 nic's each) with two storage switches. Is broadcast or unicast better? Is one more resiliant to switch failure or quicker to detect failure?
07-31-2011, 03:46 PM
To-M

Using broadcast is best used with small amount of systems on the network where unicast is best used with many servers so the management packets will be unicasted between specified the IP addresses.
08-01-2011, 10:14 PM
Olias

Quote:

Originally Posted by enealDC

There are a number of people (myself included) who are already doing this. You aren't in uncharted waters. iSCSI failover and SCSI-3 Persistent reservations has been in Open-E for a while.

Thanks for the reply. I know both features have been available, but I wasn't sure if anyone was actually using them for a redundant cluster shared volume in a Hyper-V environment. It's good to know I'm not the first...
08-02-2011, 01:23 PM
enealDC

Np. And let us know how it goes. Sharing just helps evolve the product and helps new users.
08-08-2011, 03:43 PM
marcinrozanski

how are you getting on Olias ?

You mentioned: "I've seen the documentation for using DSS in a Windows Server cluster, but not using TWO DSS storage servers to appear as a single, redundant storage device for the cluster."

That's true, one can find how to set 2 x Open-E in iSCSI failover and connect 1 x Hyper-V to it. And connecting multiple Hyper-V to Single Open-E is possible as well, BUT....

From the release notes:

“When using DSS V6 in windows 2008 cluster environment a failover event on DSS V6 will break i/o operation performed on the DSS V6 iSCSI target e.g. copying of files "

So it looks like Fully HA setup (everything x 2 ) is going to have some problems.

Apparently it's something that all storage vendors experience when dealing with Hyper-V.

I've seen guys running multiple ESX boxes off a single SAN box - failure !
SAN goes down , multiple ESX boxes loose access to the storage , 10s or 100s of VMs are down :-)
08-09-2011, 02:54 AM
enealDC

Your mileage varies with whether or not I/O gets broken, but I've tested some settings for my customers that have overcome this issue for some workloads.
I've already posted some articles on this forum that aided me..h
08-09-2011, 01:09 PM
Olias

I do have some concerns about what "DSS V6 will break i/o operation performed on the DSS V6 iSCSI target e.g. copying of files" really means in a real-world Hyper-V environmnet.

I hope to find out some time later this month. Right now I am waiting for a good time to move my current VMs off to a temporary third Hyper-V server while I build/test the cluster environment. During a catastrophic SAN failure, I could live with the VMs actually "crashing" and then restarting from the redundant SAN. 100% uptime is not a requirement for our environment; I'd just like to minimize any downtime.

Quote:

Originally Posted by enealDC

I've already posted some articles on this forum that aided me..h

I'll do some forum searches and look at your tips!
08-10-2011, 01:31 AM
enealDC

Well what it's trying to say is that when a failover occurs that there is a risk of IO operations being broken.
DSS isn't just breaking stuff randomly :)