DL580G4 and MSA1000

[Borries]

Hi,

I am having disk queue problems on one of my disk volumes.

Server is connected to MSA1000 with fiber connection. Server is running Win2003 Server Ent, 4 Quad DC CPU's, 12GB Memory; running SQL 2000 Ent Edition.

SQL Database disk config(MSA1000):
1. RAID1 146GB 15k RPM SQL Logs
2. RAID1 146GB 15k RPM SQL Indexes
3. RAID5 300GB 15k RPM(3 disks) SQL Databases

The disk queue problem is on the RAID5 disk volume. Performance monitor will give readings of around 180 (Saw bursts of more than 1000), 6 -7 times a day. The server will be very slow during this time. This can go on anything from 2 to 20 minutes. It seems pulling a report will cause the disk queue to jump but have seen the same thing on simple scripts. Will also do this when only 2 people are connected over a weekend.

We have about 200 users connecting to the server. Not a lot of writing, mostly reading.

The RAID controller has 512MB cache.

When running the HP RAID Diagnostic tool, the drives are reported at running at only 3k rpm and not 15k rpm - I think this is where the problem lies. How do I fix this?

Thanks

Johan

[Brian_Murdoch]

Johan,

If possible - connect to the MSA1000 console port on the right side controller using the MSA1000 console cable supplied with the MSA1000 and connect to the serial port of a server or laptop at 19200,8,n,1 using hyperterminal or any other decent terminal emulator.

Once you get a prompt (CLI> or CLI-1>) type in "show tech_support" This will run a script and capture all the necessary details from the MSA1000 controller(s). Capture this information in a text file. You can open a case with HP and submit this for them to investigate (along with an ADU report) or you can attach the text file to your next reply and it can be checked by forum members.

The cache can be adjusted with ACU in favour of READS or WRITES (75% READ, 25% WRITE) but your problem seems more fundamental than that. You can try it anyway to see if it helps. Open ACU and select the MSA1000. Change the controller settings to BIAS it in favour of READS and SAVE the changes. This is dynamic and doesn't require a reboot of any servers or the MSA1000 as it should flush the cache before the changes are made. The cache BIAS can be put back to 50% READ, 50% WRITE (The default) if this doesn't help.

I hope this helps.

Brian

[Borries]

Thank you, Brian.

A few things I have already done:
1. MSA1000 was replaced with new MSA1000 by HP.
2. Moved HBA from old server.
3. Moved Fiber port on MSA from old server.
4. Moved RAID controller from old MSA1000.

The "old server" mentioned here is a DL760 server, which had no problem with disk queues. RAID config on DB disk volume was RAID5 over 5 x 72GB disks.

Next step for me is to dump the RAID5 on the 300GB's and replace it with Raid1+0 over 8 x 72GB 15k RPM drives to increase the number of spindels. Do you agree with this config?

I am sure the above will not address the 3k RPM result with the diagnostig tool. I agree with you that the problem might be more serious than changing the RAID config and might not sort-out my problem. Having said that I can't belief that 2 x MSA1000's and 2 different MSA RAID controllers can give the same problem.

Regards

Johan