HP DL360 G3 Hangs on Disk Removal

[Robert Jurinac]

Hi!
I have installed Debian 3.0r2 on HP DL360 G3, with custom kernel 2.4.24. Everything worked OK until I tried to remove one of disks in RAID 1. System respondes to return key but as soon as I try to do something which needs disk it hangs (simple ls command is enough).

Here is context of /proc/driver/cciss/cciss0:
cciss0: HP Smart Array 5i Controller
Board ID: 0x40800e11
Firmware Version: 2.56
IRQ: 31
Logical drives: 1
Current Q depth: 0
Current # commands on controller: 0
Max Q depth since init: 128
Max # commands on controller since init: 128
Max SG entries since init: 31
Monitor thread period: 0
Monitor thread deadline: 0
Status: operational
Sequential access devices: 0
cciss/c0d0: 34.72GB RAID 1(0+1)

Thanks,

Robert

[Steven E. Protter]

Raid 0+1 nis not Raid 1.

Raid 1 i an exact mirror copy.

It would appear that your array is configured to stripe and mirror.

So what happened would appear to be a configuration issue.

How was it done? Disk druid? After installation change? That servers array setup needs to be redone to make it work the way you expect it too.

SEP

[Robert Jurinac]

It says "RAID 1(0+1)" which is RAID 1 (value in bracket has no efects in this case).
It is RAID 1 with 2 disk in it (36,4GB Ultra 320 SCSI 15K), configured using HPs (Compaq's:) "Option Rom Configuration for Arrays" version 2.16. And as result I have one logical drive in RAID 1, visible for Debian as /dev/cciss/c0d0.

[Roberto Polli]

Which ciss module do you have?

Being a hw raid you shouldn't see anything...

I use rh8.0a and i never experienced such a problems. Is it possible that the mirrored hd is broken?

Peace, R.

[Robert Jurinac]

cciss (drivers for the Compaq Smart Array controllers) are version 2.4.50.

What did you mean by "Being a hw raid you shouldn't see anything...".
As a hw raid I don't see both disks but just one device (which states for this two disks in RAID1) which should be OK.
Both disk are OK (same behavior is if I remove first or second)
After systems hangs, only solution is to power off/on server and than every thing works normal (but this is not the way things should behave :)

[Robert Jurinac]

It seems that firmware 2.56 for Smart Array Controller 5i+ was the problem. I upgraded to version 2.58 and now everything work OK.

[Stuart Browne]

What he meant is that hardware RAID should be tolerant enough in a RAID1 or RAID5 setup to still be able to run, and it shouldn't effect the OS.

Of which you know which is why you were here. ;)

This is a good one for the forums.. Thank you for posting your findings back here!

[Olivier Drouin]

1 question first...

Are you sure you didnt do it too fast ? The LEDs of both disk must be constant green when you take a disk... The rebuild process must be completed (so they are completely in sync) before you take a disk.

[Jeroen Peereboom]

I have firmware 2.36 (at least this is shown on the screen at boot time) and did encounter problems. Even accidentally interrupting the recovery process didn't do any harm.

JP.

[Robert Jurinac]

I have two DL380G3 with SmartArray 5i+ and firmware version 2.36 and everything is working fine, so my first idea after suspecting firmware was to downgrade it from 2.56 to 2.36. Then I find out that 2.58A is available and that one of the thing it fixes is "Fixed an issue where some ProLiant server models may become unresponsive after a drive hot-plug". Details can be found in Customer Advisory http://wwss1pro.compaq.com/support/reference_library/viewdocument.asp?source=EX031212_CW01.xml&dt=3

[Jeroen Peereboom]

Robert,

thanks for your reply.

JP