Re: linux-2.4.19-pre6 on sun4m SMP running woody
On Sun, Apr 14, 2002 at 08:55:05AM +0200, Christian J?nsson wrote:
> On Sat, Apr 13, 2002 at 01:22:30PM +0200, chj wrote:
> > I have tested the linux-2.4.19-pre6 (+ a makefile patch) on my sun4m (SS20)
> > SMP sunning debian woody for a while now. I am only aware of the swapon
> > seg fault problem but I suspect more...
> >
> > What stress tests should I do to verify any failures?
> >
Here is a stress test I used on my SS20 to verify the 2.4 kernel, ext3, and
RAID are all stable..
Run the following programs (all at the same time!) I used SSH to get into
the SS20 for all of the programs, then redirected the output of each
(using >> XXXX.log 2>&1) to a file, then SSH in again to run tail -f on the
file to track progress. In the console window (which is a serial console
as I don't connect a monitor to my sparcs) I print out the current time
every 5 seconds using a continous while loop, so that the conserver log of
the console will show exactly what time the system died if it does die.
- Start up 6 Bonnie sessions with file sizes (in MB) of: 1900, 1100, 800,
600, 400, 256. Put them in while loops so they run continously.
- Compile the 2.4 kernel
- Compile GCC 2.95.3
- Once everything has been running for several hours (before either compile
finishes, which will be *many* hours), pull the plug on the computer,
then reboot the sytsem and verify it all looks good.
- Force a check of the ext3 partitions to verify journal caught everything,
and ensure the check is running during raid resync for extra stress.
- Start up the above 6 Bonnie sessions and 2 compiles, then transfer a
large amount of data using netcat/afio.
- Pull the plug again after the file transfer has been running for awhile
- Reboot, and force e2fsck to run on all partitions.
- Start up the above 6 Bonnie sessions and 2 compiles, then let the system
just sit there for 48 hours.
If all of the above succeeds, there is a good chance the system is stable.
Another test you can run that probably is NOT good to run without swap
space is to compile the kernel (using make, not make-kpkg) with -j 5 to get
5 concurrent compiles at the same time....that puts a lot of stress on the
system, but the you have to ensure there is enough memory available or the
system will crash simply because you use up all available virtual memory,
not because it is unstable.
Good luck with your stress tests!
--David
--
To UNSUBSCRIBE, email to debian-sparc-request@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Reply to: