Re: [3.2.2] tasks blocked during matrix auto checking.

From: NeilBrown
Date: Mon Feb 06 2012 - 17:48:02 EST


On Mon, 06 Feb 2012 13:02 +0100 Pawel Sikora <pawel.sikora@xxxxxxxx> wrote:

> Hi,
>
> on heavy loaded opterons i've noticed some blocked tasks during matrix auto checking.
> is it a known issue?

No.... maybe not too surprising though.
The data-check will pause to let other IO through, but if there is lots of IO
queued up it could cause some longish delays...

2 minutes does seem a bit long though, so maybe there is a bug somewhere.

And had 3 consecutive timeouts, so that makes it 6 minutes which really is
too long.

What sort of array was this? RAID1? RAID5 ??

Thanks,
NeilBrown



>
> (...)
> [401836.109354] md: data-check of RAID array md0
> [401836.109364] md: minimum _guaranteed_ speed: 1000 KB/sec/disk.
> [401836.109368] md: using maximum available idle IO bandwidth (but not more than 200000 KB/sec) for data-check.
> [401836.109388] md: using 128k window, over a total of 8000256k.
> [401836.111441] md: delaying data-check of md2 until md0 has finished (they share one or more physical units)
> [401914.274728] md: md0: data-check done.
> [401914.293562] md: data-check of RAID array md2
> [401914.293566] md: minimum _guaranteed_ speed: 1000 KB/sec/disk.
> [401914.293569] md: using maximum available idle IO bandwidth (but not more than 200000 KB/sec) for data-check.
> [401914.293589] md: using 128k window, over a total of 849514496k.
> [402723.026480] INFO: task kjournald:1546 blocked for more than 120 seconds.
> [402723.026484] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.

Attachment: signature.asc
Description: PGP signature