Re: [Fwd: Re: mdadm: spare rebuilding]

Re: [Fwd: Re: mdadm: spare rebuilding]

[Jon Buckingham]

[-- Attachment #1: Type: text/plain, Size: 2238 bytes --]

Hi Neil

So, a little while ago I did upgrade the kernel
(2.6.22.18-0.2-default), but it did not seem to make any difference to the
'echo sync > /sys/block/md3/md/sync_action'.

After receiving your latest email, I did the following...

echo sync > /sys/block/md3/md/sync_action
(device busy reported to konsole as before)
echo idle > /sys/block/md3/md/sync_action
echo sync > /sys/block/md3/md/sync_action
(device busy reported to konsole as before)

Then checked "top" - and saw significant activity from
md3_sync and md3_raid5 processes.
Also cat /proc/mdstat showed that a resync was actively happening,
and sure enough, after a couple of hours I now have
4 active devices in the array!

So I suspect a combination of the new kernel and the "idle"
sorted the problem, whatever it was.

Thanks!
And thanks for writing mdadm and for your support.

Jon B

NeilBrown wrote:
> On Wed, July 2, 2008 4:37 am, Jon Buckingham wrote:
>> Hi Neil - any thoughts on how I can debug/fix this?
> 
> Sorry for the slow responses - I seem to have been very busy lately.
> 
> No, unmount isn't needed.
> I cannot really imagine what is happening.
> Maybe write "idle" to the sync_action file first, then try
> "sync", but that seems unlikely.
> 
> Maybe unmount, stop the array,
>    echo 0 > /sys/modules/md-mod/paramters/start_ro
> and assemble that array again
>    mdadm -A /dev/md3 /dev/sd[abcd]4
> and see if that helps.
> 
> Maybe try a newer kernel if that is an option.
> 
> NeilBrown
> 
>> Thanks
>>
>> Jon B
>>
>> -------- Original Message --------
>> Subject: Re: mdadm: spare rebuilding
>> Date: Sat, 28 Jun 2008 20:42:56 +0100
>> From: Jon Buckingham <jbuckingham@blueyonder.co.uk>
>> Reply-To: jbuckingham@blueyonder.co.uk
>> To: Neil Brown <neilb@suse.de>
>> CC: linux-raid@vger.kernel.org
>> References: <48615EE7.7000502@blueyonder.co.uk>
>> <18532.47293.808234.949005@notabene.brown>
>>
>> Neil Brown wrote:
>>> Maybe a bug that has since been fixed.
>>> What happens if you
>>>   echo sync > /sys/block/md3/md/sync_action
>>> ??
>>>
>> nas:~ # echo sync > /sys/block/md3/md/sync_action
>> -bash: echo: write error: Device or resource busy
>>
>> Should I unmount first?
>> Any other info?
>>
>> Thanks
>>
>> Jon B
>>
>>
>>
> 
> 
> 


[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/x-pkcs7-signature, Size: 3281 bytes --]

Re: [Fwd: Re: mdadm: spare rebuilding]

[Neil Brown]

On Thursday July 3, jbuckingham@blueyonder.co.uk wrote:
> Hi Neil
> 
> So, a little while ago I did upgrade the kernel
> (2.6.22.18-0.2-default), but it did not seem to make any difference to the

I was thinking more like 2.6.25.
My guess is that the fix was
   commit 25156198235325805cd7295ed694509fd6e3a29e
which is in 25-rc4.

> 'echo sync > /sys/block/md3/md/sync_action'.
> 
> After receiving your latest email, I did the following...
> 
> echo sync > /sys/block/md3/md/sync_action
> (device busy reported to konsole as before)
> echo idle > /sys/block/md3/md/sync_action
> echo sync > /sys/block/md3/md/sync_action
> (device busy reported to konsole as before)
> 
> Then checked "top" - and saw significant activity from
> md3_sync and md3_raid5 processes.

I think the resync thread was sort of "half-started".   Writing 'idle'
killed it off, then it was automatically restarted and proceeded
properly.

Thanks for reporting your success.

NeilBrown