[GE users] Jobs in wrong queues

udowaechter udo.waechter at uni-osnabrueck.de
Tue Mar 3 17:26:45 GMT 2009


Hello,
I have a strange phenomenon, where jobs run in the wrong queues (GE  
6.2U1)

First, some words about the setup:

- We have one grid
- Access to queues is managed via User's proper unix-group.
- By default no user is allowed to run any job.
- We get our users from our computing-centers ldap direcory, and 3  
primary groups exist: ustaff, uguest, ustudent

I have created a Userset: userset1 , where @ustaff, @ustudent, @uguest  
are its members and its a "acces list" and "department"
All queues have this Userset as "deny access to".

This works so far, since all of our users have one of these primary  
unix-groups.

I have defined a unix-group GE-users, where all users that are allowed  
to use the grid are members of.
A userset GE-Users exists, that has "@GE-users" as members and is  
"Access list" and "department"
A userset othergroup exists, that has "@otherunixgroup" as members and  
is "Access list" and "department"

- I have all exechosts in a "default" queue
- queueA and queueB have a disjunct set of hosts (that also are in  
default, of course)
--- queueA can be used by userset: GE-Users
--- queueB can be used by userset: othergroup

It happenes from time to time, that jobs submitted by people from GE- 
Users are executed on a host from queueB (wrong).
It also happenes from time to time, that jobs submitted by people from  
GE-Users are executed in the default-queue on the execution hosts.
Users do not define a queue when submitting jobs, since only one queue  
can be used by them anyway. I understood that the GE decided itself  
(dependign on the permissions of course) upon the appropriate queue.

Is there a problem with the "unix-group" acces stuff? I do not really  
care for the case, where one submits to queueA and the job is shown as  
being run on  "default at host" and not "queueA at host", as long as host is  
a member of queueA.
I do care about the case where one submits to queueA and the job is  
run on queueB (This is not allowed, due to funding and policy and such).

Could there be a problem with this approach in general? I have checked  
the config over and over again, and everything seems alright.

Any ideas?

Thanks,
udo.
-- 
---[ Institute of Cognitive Science @ University of Osnabrueck
---[ Albrechtstrasse 28, D-49076 Osnabrueck, 969-3362
---[ Documentation: https://doc.ikw.uni-osnabrueck.de

------------------------------------------------------
http://gridengine.sunsource.net/ds/viewMessage.do?dsForumId=38&dsMessageId=119684

To unsubscribe from this discussion, e-mail: [users-unsubscribe at gridengine.sunsource.net].



More information about the gridengine-users mailing list