Cluster will not fail over when all paths to a disk fails on active node but not inactive node.

Hi all,

We are setting up a new Server 2008 R2 cluster with two nodes. Each node is presented two LUNs from our SAN, one for the Quorum drive and another for application data.

We were testing failover today and if we removed all paths to the application data LUN for ONLY the active node but not the inactive node, we expected the cluster to failover to the inactive node since it was still able to see the LUN. Instead, however, the cluster did not fail over to node 2 instead it stayed active on the first node but simply marked the application LUN as "failed" and would not failover to the inactive node.

Does anyone know if this is expected behavior and if not, how to fix?

Thanks
LVL 1
ktpoitmAsked:
Who is Participating?

Improve company productivity with a Business Account.Sign Up

x
 
SteveConnect With a Mentor Commented:
clusters with only 2 nodes are a pain as you normally need a majority to make cluster decisions. as a majority is impossible with 2 nodes it just doesn't work properly.

have you set up an independent witness server to provide a 3rd 'vote' in the cluster?

Could you confirm what cluster config as shown in the link below you have so we can help appropriately?

http://technet.microsoft.com/en-us/library/cc731739.aspx
0
 
ktpoitmAuthor Commented:
The quorum configuration is currently "node and disk majority" and we do not have any witness servers configured. This will be the first in 3 clusters would we need a separate witness server for all three clusters or can the witness server be witness to 3 separate clusters?
0
 
SteveConnect With a Mentor Commented:
as there are only 2 nodes you will need a witness of some form to cast the deciding vote in cluster events.

I believe witness servers can generally only function for one cluster but I've never tried so cannot be sure.
0
Making Bulk Changes to Active Directory

Watch this video to see how easy it is to make mass changes to Active Directory from an external text file without using complicated scripts.

 
ktpoitmAuthor Commented:
All the nodes for these clusters are physical blades, would there be any issues with each witness server being a VM?
Also, would the LUNs need to be presented to the witness servers as well?
0
 
NetfloConnect With a Mentor Commented:
Just to add one vital question to make sure your cluster works correctly. When you created the cluster, did it pass validation? Or were the errors produced ignored?

Please see the following link for further reading: http://technet.microsoft.com/en-us/library/cc732035(v=ws.10).aspx
0
 
ktpoitmAuthor Commented:
Yes the cluster did pass validation.
0
 
SteveCommented:
All the nodes for these clusters are physical blades, would there be any issues with each witness server being a VM?
Also, would the LUNs need to be presented to the witness servers as well?

Yes it would be an issue. the witness server(s) should be independent of the cluster and should not be affected if one or more nodes fail.
0
 
ktpoitmAuthor Commented:
I've requested that this question be deleted for the following reason:

We decided to use a different method for the disks in the cluster.
0
 
SteveCommented:
Although you may not have got the answer you wanted, a valid answer was provided and the question shouldn't be deleted.
0
 
ktpoitmAuthor Commented:
I've requested that this question be closed as follows:

Accepted answer: 0 points for ktpoitm's comment #a39283819

for the following reason:

We are not going to use the witness server for cluster failover.
0
 
SteveCommented:
Does anyone know if this is expected behavior and if not, how to fix?

Responses 39281523 & 39281751 identified the issue and offered the recommended solution. Although you chose not to follow the recommendations, it was still a valid solution to your query.
0
Question has a verified solution.

Are you are experiencing a similar issue? Get a personalized answer when you ask a related question.

Have a better answer? Share it in a comment.

All Courses

From novice to tech pro — start learning today.