Home / server / Questions / 85226
In Process
Rhys Evans
Asked: 2009-11-17 09:10:43 +0800 CST

raid 5 epic fail #2

  • 772

following-

Part 1
https://serverfault.com/questions/83262/raid-5-rebuilding-epic-fail,

managed to get the raid controller half working..

when my linux distro (pmagic) boots, as far as i can tell it sees the hard disks and attempts to mount it as scsi2 but then it fails saying: 

3w-xxxx: scsi2: Command failed: status = 0xc7, flags = 0x59, unit #0
sd 2:0:0:0: WARNING: Command (0x28) timed out, resetting card.

and then just hangs there and is unresponsive to ctrl, alt + delete so probably kernel panic?

I'm assuming unit# 0 indicates there is a fault with the hard disk in position 0 on the raid controller. But surely if that was the case then would the raid controller not have let it through?

also any ideas what the flags and status are? ive tried searching google but keeps returning random errors from irrelavent sources. Cheers

raid

2 Answers

  1. You're seeing scsi timeouts on drive 0.

    a few questions:

    1. is the array degraded, or failed?
    2. what are the hard drive types, age and firmware versions?
    3. same about the controller itself

    There are a few courses of action to take here:

    1. if the array is working, but degraded, upgrade the HDD's firmware versions and the controller (driver first, then firmware), and try to rebuild. SCSI timeouts are often caused by buggy firmware on disks and controllers, and not by actual errors. If the rebuild fails - replace the disk
    2. if the array is failed, you can try to re-tag the array (clear controller configs, insert the disks, create a new raid5 similar to the old one, and don't let the controller reinitialise the new array. then pull the disk that failed first offline manually). This will tag the drives as belonging to the array, and might bring the array online in a degraded state, then use option 1.
    3. If the data is not important (i.e. you have a backup), you can start testing the disk and controller, by connecting the disk to another connector on the controller, to verify it's not the cable/connector misbehaving here. or you can insert a known good drive to test the connector instead - it's all up to you
    4. reseating the controller might also help
    5. if it's a brand name server, you can call support, you'll probaby hear what I have already said, but the support guys should have links to drivers and firmware handy, and should be aware of current hot issues
    • 2

  2. Yes Unit 0 is the one showing problem(s)

    A list of the SCSI flags http://docs.hp.com/en/A5159-96003/apas01.html

    • 0