[GE users] Listing compute nodes assigned to completed job? / New "accounting" script

Mike Hanby mhanby at uab.edu
Tue Sep 30 21:20:50 BST 2008


Thanks to both of you for the information, also, thanks for the script
Reuti.

-----Original Message-----
From: Reuti [mailto:reuti at staff.uni-marburg.de] 
Sent: Tuesday, September 30, 2008 13:30
To: users at gridengine.sunsource.net
Subject: Re: [GE users] Listing compute nodes assigned to completed job?
/ New "accounting" script

Hi,

Am 30.09.2008 um 18:59 schrieb Mike Hanby:

> Is it possible using qacct or some other tool to list the compute 
> nodes assigned to a completed job (successfully completed, error'd, 
> qdel'd, etc) aside from relying on the users log files?
>
> qacct -j JOBID reveals the master node, but none of the workers.

as Mark mentioned, it's stored in the accounting file. Besides this, you
can also find it when you request "qacct -j <jobid>". With a Tight
Integration, every "qrsh -inherit ..." call would get an entry in the
accounting file, and you would also need to add them up to get the right
timings in total for a parallel job.

Although this will list the nodes, you might have to look at the used
CPU times to check, whether you got an uneven distribution, i.e.  
something like:

node01 2
node02 1
node03 1

Please find enclosed a modified version of a script which I posted
already some time ago. By running it with "./accounting -j <jobid>"  
you will get the used hostnames and each times besides the sum.

-- Reuti


---------------------------------------------------------------------
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