[GE users] array jobs mess up fair share ??

Chris Rudge chris.rudge at astro.le.ac.uk
Fri Jul 4 11:06:48 BST 2008


Hi,

One of our users has recently started making use of array jobs on our
cluster. These are working really well but since then, the fair share
usage statistics appear to have been completely messed up. 

The accounting logs report a breakdown by project which matches the
observed usage. For our top 3 user groups/projects cluster cpu use is
split:
tag	67%
wasp	23%
eos	9%
over the last 3 months

Our fair share has a fairly long half life (90 days) and I'd say that
the share of cluster usage hasn't changed significantly over the last
year or so. If I browse through qmon --> policy configuration --> share
tree policy I see usage of
tag	24%
wasp	 7%
eos	68%

The user submitting the array jobs is in the eos group. Prior to the use
of array jobs the fair share usage matched the accounting information.

Is it possible that the fair share usage incorrectly reports usage as if
all array tasks are running from when the first one starts until the
last one finishes. For example, if each of 100 tasks takes an hour to
complete but it takes two days for all tasks to be run, is the fair
share usage reported as 100*2days rather than 100*1hour.

Is this a know issue? If so, is there any prospect of a fix (current
version is 6.1u3)?

Regards,
Chris


-- 
Dr Chris Rudge
chris.rudge at astro.le.ac.uk

Research Computing Manager
Dept of Physics & Astronomy
University of Leicester
LE1 7RH

web.  www.ukaff.ac.uk
Tel.  +44 (0)116 2523331
Fax.  +44 (0)116 2231283
Mob.  +44 (0)794 1379420


---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
For additional commands, e-mail: users-help at gridengine.sunsource.net




More information about the gridengine-users mailing list