Manually failing a disk during a md raid6 reshape

Manually failing a disk during a md raid6 reshape

[EJ Vincent]

Greetings,

I currently have a md raid6 reshape (growth, adding a disk) under-way.  
Dmesg is beginning to fill up with:

[23661.913556] ata15.00: status: { DRDY }
[23661.914443] ata15.00: failed command: WRITE FPDMA QUEUED
[23661.915326] ata15.00: cmd 61/80:f0:00:34:e9/00:00:45:00:00/40 tag 30 
ncq 65536 out
[23661.915329]          res 40/00:00:00:00:00/00:00:00:00:00/00 Emask 
0x4 (timeout)
[23661.917110] ata15.00: status: { DRDY }
[23661.917977] ata15: hard resetting link
[23664.116087] ata15: SATA link up 3.0 Gbps (SStatus 123 SControl 0)
[23664.118667] ata15.00: configured for UDMA/100
[23664.118685] ata15.00: device reported invalid CHS sector 0

What's the best way to identify ata15.00 and is it permissible to 
manually fail this device, without damaging and/or interrupting the 
reshape?

I'm also perfectly OK with waiting for it to complete on it's own, 
although the speed of the reshape operation has diminished, by my 
estimate, about 25% since beginning.

Also, any conjecture on what might lead to the above error(s)?

Thank you very much.

Best,
-EJ

Re: Manually failing a disk during a md raid6 reshape

[Mikael Abrahamsson]

On Tue, 4 Dec 2012, EJ Vincent wrote:

> What's the best way to identify ata15.00 and is it permissible to 
> manually fail this device, without damaging and/or interrupting the 
> reshape?

using "iostat -x 5" you might see the AWAIT or similar values shoot up on 
one of the drives when this is happening, that might be a good indicator. 
Apart from that it's not that easy to go from ataXX -> sdX format.

-- 
Mikael Abrahamsson    email: swmike@swm.pp.se

Re: Manually failing a disk during a md raid6 reshape

[Jack Wang]

2012/12/5 EJ Vincent <ej@ejane.org>:
> Greetings,
>
> I currently have a md raid6 reshape (growth, adding a disk) under-way.
> Dmesg is beginning to fill up with:
>
> [23661.913556] ata15.00: status: { DRDY }
> [23661.914443] ata15.00: failed command: WRITE FPDMA QUEUED
> [23661.915326] ata15.00: cmd 61/80:f0:00:34:e9/00:00:45:00:00/40 tag 30 ncq
> 65536 out
> [23661.915329]          res 40/00:00:00:00:00/00:00:00:00:00/00 Emask 0x4
> (timeout)
> [23661.917110] ata15.00: status: { DRDY }
> [23661.917977] ata15: hard resetting link
> [23664.116087] ata15: SATA link up 3.0 Gbps (SStatus 123 SControl 0)
> [23664.118667] ata15.00: configured for UDMA/100
> [23664.118685] ata15.00: device reported invalid CHS sector 0
>
>
> Also, any conjecture on what might lead to the above error(s)?

If you do not enable async scan, you may find the ataxx to sdx mapping
from looking into dmesg.

the error seems because disk not respouse ncq write command, lead to
command timeout by upper layer, which lead to error handler action.

I gusse the reason is the faulty disk not return respounse in time.

Jack