Compaq Proliant 3000 and HP NetRAID (AMI MegaRAID) with Windows 2000

[TheDameon]

I have a Compaq Proliant 3000 (dual 333Mhz and 512MB RAM; BIOS E39 dated 04/06/1998) with an HP NetRAID SCSI adapter. I currently have 2 x 16MB (32MB) cache on the card, but have tried 1 x 4MB and 2 x 4MB with no changes. I also ensured I was using the newest BIOS on the NetRAID (4.03). Not sure about the 3000 BIOS revision. There are 4 IBM 18GB drives installed internally. The drives are setup in a single RAID5 array. The system runs fine with no error messages. I am having an issue with disk access delays. During the initial Windows 2000 Advanced Server install, it (as you all know) copies files here and there. While doing this, it starts REALLY fast and then (within a few seconds) STOPS. Nothing. No screen activity, no disk activity. Then, in a few more seconds, starts again. These delays can also be much more significant (sometimes minutes) for no apparent reason. I suspected Windows 2000 as the problem, so I wiped the system and installed Linux. Same thing. Heavy disk I/O results in these odd pauses. So I went back to installing Windows 2000. After getting Windows 2000 installed, it always gives the startup error "At least one service or driver failed during system startup". Upon checking the Event Viewer, I see "Event ID: 9" "The device, \Device\Scsi\mraid35x1, did not respond within the timeout period.". Everything seems to be working fine, with the exception of these odd delays.

Thanks for your input.

 

[diamondlod]

It sounds like you could possibly (translation: this is a guess) have a cache handler problem on the NetRAID card.

It could be that the cache is filling up and then having a hard time dumping it off to disk.

You might want to try disabling cache acceleration.

 

[TheDameon]

Good guess... it was also one of my first guesses as that's what it seemed to be. I disabled caching and it did the same thing.

All is not lost though. I do have a solution and this isn't one that can be fully explained.

On a whim, I pulled one of the four disks in the RAID5 array. I then ran some tests and everything was FAST. Very fast. No delays. So I thought maybe the disk was bad in some manner and I just happened to pull the bad disk. I also though maybe the array was corrupt. I wanted to test the second theory and put the disk back into the array. It rebuilt the drive and everything is still fine.

I'm not fully sure what the problem was, especially considering I rebuilt the array twice. When I originally had the problem, I suspected a RAID inconsistency and rebuilt the array. It did the same thing so I ruled that out.

It's working fine now.

 

[DAlzamora]

Have you had any issues since?...
I'm getting the same errors in the logs.
I've had users complaining of intermitten slowness during file saves.

 

[TheDameon]

I did run some more tests and found the root of the problem. It seems that the NetRAID, Compaq, or Windows 2000 (I suspect the NetRAID), didn't like the drives I had installed. The drive I was using to test with was a 36GB drive. In a RAID5 array of 4 drives, I was swwapping one of the 18's out for a 36 and that's where the problem lies with perfect consistency. I can replicate the issue every time. Although it shouldn't matter what size I put in (as long as it is minimally the size of the other drives in the array group), it does in this case.