RAID1 over RAID0 misbehavior

RAID1 over RAID0 misbehavior

[Måns Rullgård]

I recently created a RAID1 mirror from two identical RAID0 arrays
under Linux 2.6.0.  I was surprised when /proc/mdstat reported a
resync speed of 1 MB/s, even though the system was otherwise idle.  I
increased the min sync rate in /proc/sys/dev/raid/speed_limit_min to a
large value, and was happy to see the resync rate increase to 40 MB/s.
It seems to me like something confused the RAID1 layer into thinking
there was lots of IO going on in the system.  My guess is that the
disk IO generated by the RAID0 is being overlooked by RAID1 as being
caused by the resync rather than other system activity.  Whatever the
cause, the current behavior is no good.  I have no idea how to fix it,
though.

-- 
Måns Rullgård
mru@kth.se

-
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

RE: RAID1 over RAID0 misbehavior

[Guy]

You did not state your max speed setting.
Did you try to increase the max speed?


-----Original Message-----
From: linux-raid-owner@vger.kernel.org
[mailto:linux-raid-owner@vger.kernel.org] On Behalf Of Måns Rullgård
Sent: Tuesday, January 06, 2004 7:37 PM
To: linux-raid@vger.kernel.org
Subject: RAID1 over RAID0 misbehavior


I recently created a RAID1 mirror from two identical RAID0 arrays
under Linux 2.6.0.  I was surprised when /proc/mdstat reported a
resync speed of 1 MB/s, even though the system was otherwise idle.  I
increased the min sync rate in /proc/sys/dev/raid/speed_limit_min to a
large value, and was happy to see the resync rate increase to 40 MB/s.
It seems to me like something confused the RAID1 layer into thinking
there was lots of IO going on in the system.  My guess is that the
disk IO generated by the RAID0 is being overlooked by RAID1 as being
caused by the resync rather than other system activity.  Whatever the
cause, the current behavior is no good.  I have no idea how to fix it,
though.

-- 
Måns Rullgård
mru@kth.se

-
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html


-
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: RAID1 over RAID0 misbehavior

[Måns Rullgård]

"Guy" <bugzilla@watkins-home.com> writes:

> linux-raid-owner@vger.kernel.org wrote:
>
> > I recently created a RAID1 mirror from two identical RAID0 arrays
> > under Linux 2.6.0.  I was surprised when /proc/mdstat reported a
> > resync speed of 1 MB/s, even though the system was otherwise idle.  I
> > increased the min sync rate in /proc/sys/dev/raid/speed_limit_min to a
> > large value, and was happy to see the resync rate increase to 40 MB/s.
> > It seems to me like something confused the RAID1 layer into thinking
> > there was lots of IO going on in the system.  My guess is that the
> > disk IO generated by the RAID0 is being overlooked by RAID1 as being
> > caused by the resync rather than other system activity.  Whatever the
> > cause, the current behavior is no good.  I have no idea how to fix it,
> > though.
>
> You did not state your max speed setting.
> Did you try to increase the max speed?

The max speed was 200000, and I left it there.  That is also the value
I set min speed to, just to make sure I'd max out the disks.

-- 
Måns Rullgård
mru@kth.se

-
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: RAID1 over RAID0 misbehavior

[Jose Luis Domingo Lopez]

On Wednesday, 07 January 2004, at 01:37:20 +0100,
Måns Rullgård wrote:

> I recently created a RAID1 mirror from two identical RAID0 arrays
> under Linux 2.6.0.  I was surprised when /proc/mdstat reported a
> resync speed of 1 MB/s, even though the system was otherwise idle.  I
>
Anyways, it seems wiser to make two RAID1 arrays and then a RAID0 from
them two. Apart from (maybe) preventing the problem you are
experiencing, this should give you better results in the case of a drive
failure, when it happens.

Think for a moment about the event of a single hard drive failure: with
two RAID1 arrays, one of then will go to degraded mode, but as far as I
know the upper RAID0 will not notice a thing, so will continue working
as usual, maybe with a little performance decrease.

On the other hand, if you have two RAID0 arrays and one of the hard
drives fail, the RAID0 where it happens will fail, the upper RAID1 will
run degraded, and when a new disk replaces the failed one you will have
to start the failed RADI0 again, and let the upper RAID1 reconstruct
itself. I think this situation is worse than the one depicted before.

Am I missing something ?

Greetings.

-- 
Jose Luis Domingo Lopez
Linux Registered User #189436     Debian Linux Sid (Linux 2.6.1-rc2)
-
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: RAID1 over RAID0 misbehavior

[Måns Rullgård]

Jose Luis Domingo Lopez <linux-raid@24x7linux.com> writes:

> On Wednesday, 07 January 2004, at 01:37:20 +0100,
> Måns Rullgård wrote:
>
>> I recently created a RAID1 mirror from two identical RAID0 arrays
>> under Linux 2.6.0.  I was surprised when /proc/mdstat reported a
>> resync speed of 1 MB/s, even though the system was otherwise idle.  I
>>
> Anyways, it seems wiser to make two RAID1 arrays and then a RAID0 from
> them two. Apart from (maybe) preventing the problem you are
> experiencing, this should give you better results in the case of a drive
> failure, when it happens.
>
> Think for a moment about the event of a single hard drive failure: with
> two RAID1 arrays, one of then will go to degraded mode, but as far as I
> know the upper RAID0 will not notice a thing, so will continue working
> as usual, maybe with a little performance decrease.
>
> On the other hand, if you have two RAID0 arrays and one of the hard
> drives fail, the RAID0 where it happens will fail, the upper RAID1 will
> run degraded, and when a new disk replaces the failed one you will have
> to start the failed RADI0 again, and let the upper RAID1 reconstruct
> itself. I think this situation is worse than the one depicted before.
>
> Am I missing something ?

I had a few reasons:

- I read somewhere that RAID0 over RAID1 didn't work.  Maybe this
information was outdated.
- I was short on spare disks, so I had to build an initially degraded
array and then add the other half of the mirror.
- Performance after a disk failure isn't a big issue with this
machine.  Resyncing takes only an hour or two anyway.

-- 
Måns Rullgård
mru@kth.se

-
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html