Active/Passive Slow Failover

[resoremix]

I have two DELL servers connected to a DELL Powervault SCSI storage device (7 disks) and the entire MSCS is working fine.

BUT

When I move the groups (of which I have three) to the passive node it takes approximately 2 and a half minutes for the change to take place during which time the cluster is unavailable. When the move is complete the newly active node works fine.

My question is should there be this delay or should it be a seamless transition with the cluster being available at all times?

If it is any help, the disk resources move almost instantly whereas other resources such as the cluster name are the last to move and take the full 2 and a half minutes.

I'm sure on one of my test set-ups that the move was instant.

Any help would be really appreciated.

Thanks

 

[northernrock]

Latency time ranges from seconds to minutes. Two and a half minutes is not unreasonable although it may seem like a life time when your are waiting and hoping that things all come together. Something to consider is to make sure that the standby server is equal to ( or better than but not likely )the master server in performance. It can take some time because the change over occurs when the heartbeat between the two servers stops and is detected by the standby. This is realy where most of the delay time comes into play.

 

[CourtLNet36]

Hey resoremix, I hate to hijack your thread, but are both of your Dell's seeing the one PowerVault as a single storage source? If so, how did you set that up? How did you cable the two servers to the PowerVault? I've been trying to get my cluster set up like that.

 

[resoremix]

CourtLNet36

Yes they do see the Powervault as one storage source. The cabling is straight forward, just one SCSI cable from each server to one of the SCSI interfaces on the Powervault.

So.. Your problem will lie in one of two places. Firstly make sure that the SCSI interfaces in your Powervault are what Dell refer to as the 'clustering option'. The two interfaces that are supplied as standard are not designed for cluster use as they effectively act as two physical / two logical interfaces rather than the required two physical / one logical. The clustering interfaces will have a jumper on them that you must ensure is set to 'cluster'.

If you're OK on the Powervault's interfaces then the other likely source of your problem is the SCSI boards in the two servers. I don't know what servers / interfaces you are using but in my case it was Dell Poweredge 1400's. All I can say is that once again the SCSI interfaces you use on the server must be 'cluster aware'. Dell supply what they call Power Edge Raid Controllers (PERC) and we are literally talking about the only visible difference (aside from cost ;-)) between a normal controller and a cluster capable controller being a couple of lines in the card's BIOS.

So I would check all of your interfaces first. RAID capable is not good enough I'm afraid. They need to be cluster capable as well.

I found these these links invaluable:

http://www.microsoft.com/technet/tr...tions/ecommerce/maintain/operate/d5clustr.asp

http://www.microsoft.com/windows2000/techinfo/planning/server/clustersteps.asp

Hope this helps.