Clustered SQL Servers and chkdsk?

Hi all,
We have 2 SQL Server 2000's in a cluster at work, we started running some
regular server maintenance a few ws back including the basics like defrag
etc...one of the things I was also keen to do was to run chkdsk /F after we
had done the defrags and reboot the servers, I assumed we'd be able to do
this on the passive server, then when it was back up, fail over the active
server and then run it on that one also. Of course, what I failed to
consider was the fact that the shared E drive is a resource which only
'exists' on the active server, so when I fail the active server over, the E
drive goes with it.
From memory the E drive is raid'ed and a colleague mentioned that because of
this there might not be any benefit running a chkdsk anyway, I was hoping to
confirm this, or find a way of chkdsk'ing the shared resource - the only
option I can see at present is to dismount the drive and run it on the
active server but this would then obviously involve some down time, plus
I've never dismounted a drive, and certainly not one being used excessively
like ours on the SQL Server, as a result I have no idea of the consequences.
Any advice would be appreciated - I was previously of the understanding that
running defrag on drives that have large quantities of I/O requests was a
good thing to help reduce the movement of the head in the disc, thus
inproving performance (if only slightly). I assumed chkdsk would be another
worthy program to run (as we do on the other servers, web and the like).
Regards
RobWhen you run chkdsk you're reordering the data on the physical disk. This
would help for most file types. However, I don't think you would gain much
if the drive is striped. On also, on db creation, the file is preallocated
(i.e. taken up whatever the specified size). If the disk is fragmented at
that time, the file is already fragmented. You might gain some perf if you
take the database offline and do chkdsk.
Yup, without a true shared clustered filesystem (i.e. with MSCS/Veritas/etc)
the shared disk is under control of the active node.
Below is a very good whitepaper on fragmentation and what you can do in
regards to sqlserver. I highly recommend a good read.
http://www.microsoft.com/technet/pr...n/ss2kidbp.mspx
-oj
"Rob Meade" <robb.meade@.NO-SPAM.kingswoodweb.net> wrote in message
news:OzCId.16421$GG1.8815@.text.news.blueyonder.co.uk...
> Hi all,
> We have 2 SQL Server 2000's in a cluster at work, we started running some
> regular server maintenance a few ws back including the basics like
> defrag
> etc...one of the things I was also keen to do was to run chkdsk /F after
> we
> had done the defrags and reboot the servers, I assumed we'd be able to do
> this on the passive server, then when it was back up, fail over the active
> server and then run it on that one also. Of course, what I failed to
> consider was the fact that the shared E drive is a resource which only
> 'exists' on the active server, so when I fail the active server over, the
> E
> drive goes with it.
> From memory the E drive is raid'ed and a colleague mentioned that because
> of
> this there might not be any benefit running a chkdsk anyway, I was hoping
> to
> confirm this, or find a way of chkdsk'ing the shared resource - the only
> option I can see at present is to dismount the drive and run it on the
> active server but this would then obviously involve some down time, plus
> I've never dismounted a drive, and certainly not one being used
> excessively
> like ours on the SQL Server, as a result I have no idea of the
> consequences.
> Any advice would be appreciated - I was previously of the understanding
> that
> running defrag on drives that have large quantities of I/O requests was a
> good thing to help reduce the movement of the head in the disc, thus
> inproving performance (if only slightly). I assumed chkdsk would be
> another
> worthy program to run (as we do on the other servers, web and the like).
> Regards
> Rob
>|||"oj" wrote ...

> When you run chkdsk you're reordering the data on the physical disk. This
> would help for most file types. However, I don't think you would gain much
> if the drive is striped.
Hi Oj,
I believe our RAID'd discs are raid5 if that makes a difference? (ie, not
striped)

> Yup, without a true shared clustered filesystem (i.e. with
MSCS/Veritas/etc)
> the shared disk is under control of the active node.
ta for the confirmation.

> Below is a very good whitepaper on fragmentation and what you can do in
> regards to sqlserver. I highly recommend a good read.
>
http://www.microsoft.com/technet/pr...n/ss2kidbp.mspx
Cheers, will take a read-ee now :o)
Regards
Rob