Re: 2-node cluster won't stop sqlservr.exe during user-initiated failo

Any clues in the cluster host nodes' application event logs?

--
Geoff N. Hiten
Senior Database Administrator
Microsoft SQL Server MVP

"David S." <DavidS@xxxxxxxxxxxxxxxxxxxxxxxxx> wrote in message
news:D17A3996-4321-4A97-A708-E9E6F6E25B88@xxxxxxxxxxxxxxxx
>I have the following cluster config:
>
> 2x HP DL585 servers (identical) - 4x Opteron 852 processors, 8GB RAM,
> Win2003Enterprise, SQL 2000 Enterprise
> 1x HP MSA 500 SCSI shared-storage
> 1x HP SCSI adapter per server (included with MSA500, not sure of the model
> #)
>
> The servers are in an active/passive configuration. Either one works fine
> during normal operation. But when we do a "move group" for the SQL Server
> group from either node to the other, the move process stalls waiting for
> sqlservr.exe to stop.
>
> These are pretty fast machines, but we've waited 20 minutes, with almost
> no
> CPU activity, before killing sqlservr.exe in task manager. At that point
> the
> cluster resources complete their move to the passive node and everything
> starts up just fine. We don't experience any database corruption, though
> we've de-activated all the db client applications before doing this, so
> the
> db server load is very near zero.
>
> My suspicion is that this is related to roaming profiles (required for
> encrypting file system with Win2003 clusters), which are located on the
> same
> machine in a cluster share, in the SQL Server cluster group, with SQL
> Server
> dependant on the share.
>
> Any thoughts?
>
> -David S.


.