lrkwalkers
asked on
Windows 2003 Disk Queue at 100%
In Performance Monitor, the Average Disk Queue is flatlining at 100%. Obviously, users are complaining about slow response times when accessing data.
Is there a quick way to find out which process is hammering the disks?
I've restarted the SQL services and print spooler.
Is there a quick way to find out which process is hammering the disks?
I've restarted the SQL services and print spooler.
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
" "Avg. Disk Queue Length" is a measure of the items waiting to be written into the disk. It is displayed as number of write operations in the cue, not percent. "
Should have read:
" "Avg. Disk Queue Length" is a measure of write operations waiting at the disk to be written. It is displayed as number of write operations in the cue, not percent. "
Not "Items" but "Write Operations"
Sorry.
Should have read:
" "Avg. Disk Queue Length" is a measure of write operations waiting at the disk to be written. It is displayed as number of write operations in the cue, not percent. "
Not "Items" but "Write Operations"
Sorry.
More likely than not the bottleneck is caused by SQL and your disk layout. I have listed some links below on monitoring the server and best practices:
http://sqljunkies.com/Article/D1B7C756-4725-4D31-A53D-C0A47976E6BB.scuk
http://msdn2.microsoft.com/en-us/library/ms175903.aspx
http://www.microsoft.com/technet/prodtechnol/sql/2000/maintain/sqlops6.mspx
http://sqljunkies.com/Article/D1B7C756-4725-4D31-A53D-C0A47976E6BB.scuk
http://msdn2.microsoft.com/en-us/library/ms175903.aspx
http://www.microsoft.com/technet/prodtechnol/sql/2000/maintain/sqlops6.mspx
Avg. Disk Queue Length is the average number of both read and write requests that were queued for the selected disk during the sample interval.
Generally you want to seperate your log files from your database files and you do not want to put the logs on a RAID5, use RAID10 instead. Any RAID setup will work for the database files.
Generally you want to seperate your log files from your database files and you do not want to put the logs on a RAID5, use RAID10 instead. Any RAID setup will work for the database files.
ASKER
Points awarded.
Solution in the end:
We have GFI Mail Archiver and we recently put another Exchange server into our domain for another site. Since GFI Mail Archiver wasn't licensed for all the extra users, we are assuming that the email data was trying to archive but couldn't due to the lack of licenses. Once we took the email archive databases offline, the disk queue went back to normal.
Solution in the end:
We have GFI Mail Archiver and we recently put another Exchange server into our domain for another site. Since GFI Mail Archiver wasn't licensed for all the extra users, we are assuming that the email data was trying to archive but couldn't due to the lack of licenses. Once we took the email archive databases offline, the disk queue went back to normal.
ASKER
Why is it recommended that the log files be separated from the database files? Do you mean on another RAID array or logical drive?
Different physical disk (be it RAID or not). Transaction logs are accessed sequentially, database files are random. If DB files are located on different disks then you have parallel I/O.
I've experienced this a few times and the AntiVirus scanner was causing this, even in normal realtime protection, changing priority settings helped. When experiencing heavy disk I/O remember to check the AV also!
ASKER