need help troubleshooting software raid failures

[Eric Shubert]

Alex Dean wrote:
> 
> On Jul 16, 2009, at 12:47 PM, Eric Shubert wrote:
> 
>> Alex Dean wrote:
>>>
>>> On Jul 16, 2009, at 11:08 AM, Eric Shubert wrote:
>>>
>>>> I'd look into the controller. The Promise PCI SATA cards are known 
>>>> to be
>>>> problematic with linux. I don't know the details, but have seen 2 of
>>>> them have problems with software raid at least. The problem can be very
>>>> sporadic, so it seems to fit your symptoms. Google for the details.
>>>>
>>>> Please let me know if/when you find a PCI/SATA card that works!
>>>
>>> I followed the emails about problems with Promise controllers a few
>>> months ago.  I'd planned to buy a Promise card up to that point, but
>>> went with a SYBA card w/ a SIL3112 chipset, since it seemed to be better
>>> supported / less buggy in linux.
>>>
>>> This is the one I have:
>>> http://www.newegg.com/Product/Product.aspx?Item=N82E16815124006
>>>
>>> alex
>>>
>>
>> Hmmm. I'm betting it's a hardware issue of some sort. Is it always the
>> sda drive that drops out?
> 
> Yes, at the moment 3 of the 4 sda partitions are failed.  By hardware, 
> do you mean the disk or the controller?

Yes. Neither in particular. Could be either.

> Any idea why this would start 
> occurring after a power loss?

Only an uneducated guess, which I won't bother sharing. ;)

> It's not much money to replace the controller.  The disk also wouldn't 
> be too bad, and probably is in warranty.  But I'd like to actually 
> verify it's bad before going down that road.
> 
> If there are no other good PCI SATA cards out there, I'd need a new 
> motherboard or new machine, since PCI is all I have in this computer.  
> (It's rather old...  perfect for the job I've given it.)

I went a little different route on my backup server, thinking "outside 
the box". ;)

I too re-provisioned an older machine, but am using USB drives, also 
with software raid1. The only trouble I had was that I needed to add a 
"sleep" in the init script because it would try to assemble the raid 
devices before the USB drives were up. The sleep gives the USB drives 
the little extra time they need to become active. I know it would be 
better to put in a wait loop, but it works ok as is.

If you're thinking of doing this, I'd be sure to have USB2.0. It takes 7 
hours or so to resync a 500G raid1 mirror with USB2 (my 80G pair takes 
about an hour). I wouldn't want to wait twice that long on USB1.

> Can anyone read anything more from the syslog entries and other info I 
> posted in my original message?

I looked at them in a little more detail, and my impression is like I 
said, a hardware problem. Controller or drive.

I suppose you could take the whole sda offline and run some drive 
diagnostics on it. I would think that 'shred' would show any problem 
with it. I imagine that there are other tools you could use. If the 
drive's ok, then you can add it back into the array, and you'll be 
pretty certain that the controller's flakey.

BTW, do you know for sure that your card had a different chipset than 
the promise cards?

> alex
> 


-- 
-Eric 'shubes'