Extremely high Average Disk Queue Length in perfmon?

Hi All

Im trying to diagnose a slow running issue with a ProLiant ML370 G2 W2K
(SP4) file and print server, the main complaint from users is long delays in
opening / saving documents (30-60 sec to open a small word doc for example)
and slow running of databases.
Ive been searching for the root cause and noticed that the Average Disk
Queue Length perfmon graph goes absolutely haywire on both drive arrays,
peaking up in the millions(!) for a second or two then immediately coming
back down again to zero, then repeating this continuiously. From what Ive
read Microsoft say this queue should generally be between 0 and 10 (1.5x the
number of HDD spindles in the array?) and this is the behaviour Ive seen on
our other servers, so I do believe that something is going wrong somewhere on
this server.

Because its doing it on both arrays (c: logical drive is 2x9Gb mirrored, d:
is 4x36Gb RAID5), we concluded it may be an issue with the RAID controller so
had that swapped out, but that has not helped.

It still smells of a hardware issue to me but if it is, it must be further
up the chain than the RAID controller assuming one disk fault would not cause
this on BOTH arrays? So Before we swap anything else out I was hoping that
you guys may be able to give some inspiration on what may be causing this or
things to try?

Sorry for cross posting this from the file system / services forum but after
posting it, I realised this was probably a more relevent forum.

thanks

Chris

"ChrisG" <ChrisG@discussions.microsoft.com> wrote in message news:5579BB75-CB35-472F-B6B5-D448D928EFFF@microsoft.com...
> Hi All
>
> Im trying to diagnose a slow running issue with a ProLiant ML370 G2 W2K
> (SP4) file and print server, the main complaint from users is long delays in
> opening / saving documents (30-60 sec to open a small word doc for example)
> and slow running of databases.

> Ive been searching for the root cause and noticed that the Average Disk
> Queue Length perfmon graph goes absolutely haywire on both drive arrays,
> peaking up in the millions(!) for a second or two then immediately coming
> back down again to zero, then repeating this continuiously. From what Ive
> read Microsoft say this queue should generally be between 0 and 10 (1.5x the
> number of HDD spindles in the array?) and this is the behaviour Ive seen on
> our other servers, so I do believe that something is going wrong somewhere on
> this server.
>
> Because its doing it on both arrays (c: logical drive is 2x9Gb mirrored, d:
> is 4x36Gb RAID5), we concluded it may be an issue with the RAID controller so
> had that swapped out, but that has not helped.
>
> It still smells of a hardware issue to me but if it is, it must be further
> up the chain than the RAID controller assuming one disk fault would not cause
> this on BOTH arrays? So Before we swap anything else out I was hoping that
> you guys may be able to give some inspiration on what may be causing this or
> things to try?
>
> Sorry for cross posting this from the file system / services forum but after
> posting it, I realised this was probably a more relevent forum.
>
> thanks
>
> Chris
>
>

I assume there were meant to be some of your own words on the end of that
post Glen? :)

"ChrisG" <ChrisG@discussions.microsoft.com> wrote in message news:08AAEEF5-70E3-4ED9-AFFA-1C6107DADB66@microsoft.com...
> I assume there were meant to be some of your own words on the end of that
> post Glen? :)

Sorry, yeah. Just was gonna ask if you're running Novell
client software on your workstations. Long delays in open/
save dialog boxes was a known issue with some versions of
the Netware client.

Thanks Glen, yep Ive read about the Novell Client issues but we don't run it
on our workstations. Its possible Im chasing a red herring with this perfmon
graph and it may just be a perfmon issue and not an actual fault with the
server, but Ive looked into the other common causes of client slowdown such
as oplocks etc. Any other ideas?

Cheers

Chris

"Glen" wrote:

> "ChrisG" <ChrisG@discussions.microsoft.com> wrote in message news:08AAEEF5-70E3-4ED9-AFFA-1C6107DADB66@microsoft.com...
> > I assume there were meant to be some of your own words on the end of that
> > post Glen? :)
>
> Sorry, yeah. Just was gonna ask if you're running Novell
> client software on your workstations. Long delays in open/
> save dialog boxes was a known issue with some versions of
> the Netware client.
>
>
>

How many files are in the folders where you're saving and
opening? If it's thousands then the process will be slow.

One test would be to make sure the Indexing service is
running on a workstation, then do a search on all hard
drives and network drives for a file (can be any filename).
Then see if the open/save dialogs open faster after this
search. I'd be curious to know if it makes any difference.

"ChrisG" <ChrisG@discussions.microsoft.com> wrote in message news:3BF08DC9-2E68-4F8C-A7DF-555A3AB34481@microsoft.com...
> Thanks Glen, yep Ive read about the Novell Client issues but we don't run it
> on our workstations. Its possible Im chasing a red herring with this perfmon
> graph and it may just be a perfmon issue and not an actual fault with the
> server, but Ive looked into the other common causes of client slowdown such
> as oplocks etc. Any other ideas?
>
> Cheers
>
> Chris
>
> "Glen" wrote:
>
> > "ChrisG" <ChrisG@discussions.microsoft.com> wrote in message news:08AAEEF5-70E3-4ED9-AFFA-1C6107DADB66@microsoft.com...
> > > I assume there were meant to be some of your own words on the end of that
> > > post Glen? :)
> >
> > Sorry, yeah. Just was gonna ask if you're running Novell
> > client software on your workstations. Long delays in open/
> > save dialog boxes was a known issue with some versions of
> > the Netware client.
> >
> >
> >