Print

---

Hi,

Atlas has introduced a new parameter in schedconfig to set the size of 
the jobs working directory on the WN. As it is explained below this will 
be used as a limit and as a brokering parameter so it cannot be left 
empty. For most sites for now it has been set to the value of 
maxinputsize in panda + 2GB allocated for the output. The default value 
of maxinputsize is 14GB for most sites so maxwdir=16GB for most sites. 
If this size is not correct for your site please let us 
([log in to unmask]) know what size you would like it 
to be set to.

thanks

cheers
alessandra

-------- Original Message --------
Subject: 	MAXWDIR - a new schedconfig parameter
Date: 	Wed, 24 Oct 2012 11:35:44 +0200
From: 	I Ueda <[log in to unmask]>
To: 	atlas-adc-cloud-all (contact for all the ATLAS cloud supports) 
<[log in to unmask]>
CC: 	atlas-adc-panda-support (Support forum for the Panda system.) 
<[log in to unmask]>



Dear Clouds,

As was presented at the last ADC weekly;
   https://indico.cern.ch/conferenceDisplay.py?confId=213765

We have set up a new schedconfig parameter 'maxwdir'
to specify the maximum workdir size for each job,

This should basically be set as
   (the total disk capacity that jobs on a node can use) / (number of job slots on the node)

The parameter will be used in the next version of pilot
to limit the disk usage by each job.

Another implementation in panda/prodsys will make it possible
to broker jobs based on this parameter in the near future.


We have filled this parameter in principle as
   maxwdir = maxinputsize + 2GB

considering this would be a safe assumption,
although some sites might have set maxinputsize
in different ways, eg. (available disk per job) - 5GB

Please check the values in schedconfig and adjust it, if necessary,
according to the reality.

regards,		ueda