Why don't you limit the number of jobs running at any one time? Either
put a group limitation in or a user limitation. Grep for Johannes' DN
in the gatekeeper logs and it will tell you the pool account he's
mapped to.
"/O=GermanGrid/OU=LMU/CN=Johannes Elmsheuser"
If you get into real trouble then you can ban Johannes' certificate,
by adding it to
/opt/glite/etc/lcas/ban_users.db
Cheers
Graeme
On Wed, Apr 29, 2009 at 6:22 PM, Christopher J.Walker
<[log in to unmask]> wrote:
> Graeme Stewart wrote:
>>
>> Hi Folks
>>
>> Dan has made some optimisation to the hammercloud framework and has
>> started 3 concurrent tests in the UK (up to 450 jobs per site).
>
> Please limit the number of user analysis jobs at QMUL to significantly below
> the level at which it started to fail last time. Otherwise, time spent
> sorting that out will be taken away from time spent getting storm/lustre
> installed and working.
>
> Thanks,
>
> Chris
>
>>
>> Cheers
>>
>> Graeme
>>
>>
>> ---------- Forwarded message ----------
>> From: Daniel van der Ster <[log in to unmask]>
>> Date: Wed, Apr 29, 2009 at 3:28 PM
>> Subject: Another test in UK
>> To: Graeme Stewart <[log in to unmask]>,
>> atlas-dist-analysis-stress-testing-coord
>> <[log in to unmask]>
>>
>>
>> Hi Graeme,
>> I just created 3 meta tests for UK (to test the parallel test
>> submission again) to start at 16:01 today (another small test from
>> Mark will start at 16:00). I made a change to the ganga repository
>> type to see if this decreases the loadavg during submission.
>> Dan
>>
>>
>>
>
--
Dr Graeme Stewart http://www.physics.gla.ac.uk/~graeme/
Department of Physics and Astronomy, University of Glasgow, Scotland
DEATH TO MEETINGS!
|