hardware freezez and no clue where to start searching

[AndiSHFR <AndiSHFR@gmx.net>]

Hi.

Maybe someone can give us a hint or direct us in the right direction.

We have a debian testing system. uname -a shows:

Linux sv-vmhost02 2.6.18-4-amd64 #1 SMP Mon Mar 26 11:36:53 CEST 2007
x86_64 GNU/Linux

The machine seems to freeze every n seconds for some seconds.
It can be easily seen when doing disk io.

Here is an example:

sv-vmhost02:/home# date ; time dd if=/dev/zero of=test.dat bs=1MB
count=1000 ; date

Thu May 10 17:53:45 CEST 2007
1000+0 records in
1000+0 records out
1000000000 bytes (1.0 GB) copied, 2.89129 seconds, 346 MB/s

real    0m58.632s
user    0m0.000s
sys     0m3.480s
Thu May 10 17:54:44 CEST 2007

sv-vmhost02:/home#

The date output shows that the dd did run from 17:53:45 to 17:54:44.

But the dd command thinks it only took 2.8 seconds to create the 1 GB
file -of course with a top thruput of 346 MB/s.

But the time-command shows that the dd really took 58 seconds.

So the real thruput was around 17 MB/s

So i think the whole hardware is freezing an the os does not see it
because there are
no error messages in dmesg.

When doing a "iostat -mxtc 1" during the dd it is shown that write
thruput of the /dev/sda drops to zero for seconds (typically 10
seconds) and then rises again.

Any idea where to start searching for?

Thx
Andi