Re: raid5 resync aborted under heavy XFS use

[Seth Mos <knuffie@xxxxxxxxx>]

At 21:03 1-8-2001 -0700, Matt Ryan wrote:
> Precedence: bulk
> X-UIDL: 996725357.maildrop8.8834
>
> hi -
>
> >  o There was a stall problem where the raid would just grind to a halt,
> >    this was fixed by a kernel change in the 2.4.7-pre series. This was
>
> sorry to beat a dead horse, but I have a couple questions related to
> this and the fact that I might have to go with the redhat-based 2.4.3
> 1.0.1 release kernel on a couple of machines.
>
> - what was the fix to the 'grind to a halt' problem?  was it a patch
> against the md code in particular, or some other part of the kernel?

You would have to scour the archive for the TAKE messages of the past week.
You can then see what files have been touched and fetch diffs from the 
webcvs frontend.

> - does anybody know if this same problem existed in the redhat 1.0.1

RedHat never released a 1.0.1 (was that mothersday?).
Do you mean The RedHat 7.1 2.4.3 updated kernel?

> kernel?  I don't know how much rigorous testing that kernel has been
> subjected to.  is there a specific test that triggered this problem in
> the stock + xfs kernel that I could run on the redhat kernel?

This was a specific interaction between xfs and the md raid code IIRC. This 
would not happen with ext2 but I believe reiserfs was also affected because 
of the journaling.
The 2.4.5 based xfs kernel had the same problem and even the CVS tree 
untill last week.

> - if this problem does turn out to be in the redhat 1.0.1 kernel, and
> there is a specific, simple patch that fixes it, my next step would be
> to figure out if it has any chance of working with the 2.4.3 -ac
> kernel.  I wonder what my chances of that would be.

They are beating on a 2.4.7 based tree. No exact plans or release dates for 
1.0.2 yet.
The chances of patching a -ac kernel are slimm ;)


--
Seth
Every program has two purposes one for which
it was written and another for which it wasn't
I use the last kind.