Automatic Failover + Windows Failover Clustering

Hey!


I'm setting up a Windows Server 2008 R2 Hyper-V cluster, with iSCSI connections to an Open-E machine. I've got it all up and running, but the problems start at the automatic failover of Open-e.

I've configured 2 Open-E machines in Failover mode, with all replication tasks running and the virtual ip configured. I've connected the 2008 nodes to the iSCSI targets, formatting the volume, adding it to the cluster and creating a Virtual Machine on that clusterdisk. When I trigger the manual failover with the Virtual Machine doing a disk stress test, I can see the MB/s drop to 0. After about 10 seconds, the virtual machine crashes, saying it can't connect to the storage. In the Failover Cluster management tool I can actually see the Cluster Disks failing and directly coming back online.

I've tried setting a higher iSCSI and Disk timeout in Windows, but all the timers seem to have zero effect. When I connect directly to an Open-e server, without failover or a virtual IP, and I pull out the cable while running a VM, nothing happens. When I plug it back in, the Virtual Machine continues, as if there was nothing wrong in the first place.

It's almost as if the Cluster Service of Windows 2008 'knows' it's not the same machine who is responding to the Virtual ip...

Allthough it continues directly after crashing, the virtual machine actually do crash. And that's not my idea of a failover :)



How to solve this problem?

Thanks in advance!

Peter

it's not supported yet, so you can have failover on Open-E + single Windows / Hyper-V or you can have single Open-E + Windows / Hyper-V cluster.

From release notes of the latest update ([2011-03-01] Open-E DSS V6 ver. 6.00 up65 build 5217) :

"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"

I think that Open-E is working on this problem.

Real HA shoud have (at lest):

- 2 x Open-E in failove
- 2 x Windows in cluster
- 2 x switch
- 2 x administrator (if one will take day off :-) )

Thank you for your answer!


How about VMWare ESX server? Does Open-e support that in their Failover function? I can't really get a clear answer to that with Google..

Should be possible , I haven't seen a step-by-step manual but I've got the diagram.

Please try this donwload - https://rs317tl2.rapidshare.com/#!download|317l33|457229643|Solution_proposal_HA_C luster_with_VMware_ESX4_and_DSS_V6__Workshop_20.05 .2010_.pptx|236

Its a PP presentation - Solution proposal HA Cluster with VMware ESX4 and DSS V6 (Workshop 20.05.2010)

I'm sure I've got it of the Open-E website, but can't locate it now.

Feel free to drop me an email to : martin@broadberry.co.uk - I'll send the file back if the link doesn't work.

Please have a look at our webcast and videos:
http://www.open-e.com/library/webcasts-and-videos/

You can find there some informations about:
DSS V6 iSCSI-Failover and Multipath with ESXi4.1 - EN
Open-E DSS V6 MPIO with VMware ESX4 and Win 2008 - EN

Check also the following document:
http://documents.open-e.com/Open-E_D..._ESX4i_4.1.pdf

I think that PeterKeuning isn't looking for 2 x Open-E in failover and 1 x ESX talking to Open-E thru MPIO.

He's more looking for 2 x Open-E in failover and 2 x ESX in cluster - no single point of failure !

Yes but the ESX HA regardless will work if the the HA in ESX will be using the same VIP for its setup with or without MPIO.

I have a customer running in this configuration and we perform fail over tests every 1/2 and have no real issues. (knock on wood).
There are a couple of threads on this forum that I have contributed to that you should look at. I've found some resources online that seemed to fix the problem for my environment and they may help yours too..