Re: 2.6.9: unkillable processes during heavy IO

[Brad Fitzpatrick]

Nathan,

On Wed, 17 Nov 2004, Nathan Scott wrote:

> > > On Sun, 14 Nov 2004, Brad Fitzpatrick wrote:
> > >
> > > > We have two database servers which freeze up during heavy IO load.  The
>
> Brad, could you send me details on how you've setup mysqld
> and how to generate a load similar to yours, so that I can
> reproduce the hang locally?

The MySQL people made a tool to reproduce MySQL-like load for the specific
purpose of not putting you through the database setup pain:

http://sysbench.sourceforge.net/

To simulate our InnoDB:

   -- use the tool to do "seqwr" and write out a 50GB file or so

   -- use the tool to make another file, about 30GB.

   -- use the tool to do "rndrw" to do random I/O over that 50 GB
      file, with the O_DIRECT flag on, with a bunch of threads doing
      16k reads/writes.

   -- use the tool to do random IO w/o O_DIRECT on the smaller file
      at the same time as the O_DIRECT run, also with a number of
      threads, doing smaller reads/writes.

This is happening on a machine with LVM2, as well as directly on
/dev/sdb2 (an IBM ServeRAID), so device-mapper shouldn't be an issue one
way or another.

The filesystem size is 270 GB on one machine (on /dev/sdb2) and 120 GB on
LVM2.


> > > > The hardware/software stack is:
> > > >
> > > >   - Dual Opteron 246, SMP kernel, w/ NUMA
> > > >   - 9 GB of memory (4GB in one zone, 5GB in the other)
> > > >   - MySQL, running mostly InnoDB, but some MyISAM
>
> ( I don't even know what those two things are, so you can
> probably guess at the level of assistance I'll need here. :)

Well, sysbench should help you find the problem.

Let me know if I can help more.  Thanks for looking into this.

- Brad


>
> thanks!
>
> --
> Nathan
>
>
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/