[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[HTCondor-users] Execute jobs as a certain user / probmens being "nobody"



Hi there,
we are trying to use condor for distributed rendering of cgi, using Autodesk Maya/mental ray. I am somehow confused about the exact way condor jobs are executed.

First of all the facts:
  • CentOS 6.3 (RedHat Clone) x86_64 on all machines
  • Installing condor via condor_install script
  • Setup a condor user as proposed in the manual, login is /sbin/nologin but condor is completely installed into that users home folder (install- as well as local-dir)
  • condor user, but no group "condor", use existing group "users", specified that using CONDOR_IDS
  • runnig the deamons as root (atm still started manually)
Several different users submit jobs from their Workstations, but on the computing machines for simplicity there exist just one worker user account ("render") that atm is used to run the render jobs (and has some program settings stored locally). Data is located on a shared nfs folder. I wasn't aware that it by default condor executes jobs as the submission user, cause I did first tests locally on my machine.

If I got it right there is no way to define a different/standard user that should run jobs on a certain execute machine. According to the docs in this case jobs may run as "nobody" - so I added a run_as_owner=False to my submission file,  - with the result that there is a problem concerning write permissions in the execute directory:

Starting "/usr/autodesk/maya2013-x64/bin/maya"
Warning: temp directory /home/condor/execute/dir_5417 does not have write permissions.
Resorting to default.
*** Fatal Error: Failed creating directory: /home/felix/maya
Maya could not create the necessary startup directories.
Please check for sufficient disk space and necessary write permissions. 

The folder /home/condor/execute (or $(LOCAL_DIR)/execute) initially was owned by condor:root (why that?) and has permissions 755, changed it to condor:users - the above problem appeared in both cases. I just don't know as whom condor tries to create/access this directory.

I also tried to set the SLOT1_USER to "render", which is ok for I am using only one slot per machine (Maya cares about the multithreaded rendering). Not sure if this makes sense at all, anyway it does not change anything and the problem from above still occurs.

Thanks for any help and sorry if I overlooked something obvious here.
Felix

-- 

Dipl.-Ing. Felix Ulber

[0x1] Software und Consulting GmbH
Ulmenstraße 10
22299 Hamburg

felix@xxxxxxxxxxxxxxxx
www.0x1-software.com