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

RE: [Condor-users] A Horrible Condor Bug! :-)



> Hello All (and Condor Team),

Hi, Jeff!  Great to see you again at Condor Week.

> When I try to submit a vanilla job to our pool, I get...
> 
> WARNING: File /home/jeff/poly/porta/galaxy62/porta.out is not
writeable
> by condor.
> 
> WARNING: File /home/jeff/poly/porta/galaxy62/porta.err is not
writeable
> by condor.
> 
> The horrible bug is that "writeable" is misspelled.  :-)
:-)

> Actually, I am writing to see if someone can help me figure out how
our
> configuration is wrong and why the vanilla job will not write to the
> files /home/jeff/poly/porta/galaxy62/porta.out.
> 
> In the StarterLog, I see an entry...
> 3/25 16:12:14 Starter running as 'condor', no uid switching possible
> 
> This certainly is the crux of the problem, no?

Yes indeed.

> I didn't do the install, so I am not sure what our SysAdmin did to
have
> the Starter run as "condor".  I want it to run as root.

The SysAdmin started condor when he/she was logged in as user condor.
To get privilege switching, you have to start condor as root.  The
condor daemons will have a 'real' uid of root, but change to user condor
most of the time for safety's sake.  If the condor master was started as
user condor, then you'll always be user condor: this is what you're
seeing.

> I logged in as root and did a "condor_restart -all", but that didn't
do
> the trick -- I still get the "Starter running as 'condor'" message.

It _almost_ did the trick.  The master re-execs itself, so it still
doesn't gain root privs.

> Can someone suggest the fastest way to "fix" our installation, so that
> condor can do the requisite UID switching so things are happy in the
> vanilla universe?  Or is there something else that might be wrong.
> (UID_DOMAIN looks OK to me...)

Solution A: Make your in/out/err files writ[e]able by user condor.  
Solution B: Ask the sysadmin to start condor as user root everywhere.
Solution C: (Assuming the sysadmin won't have a problem with this) 
#1 condor_config_val master  (to find out where the condor master is 
   located)
#2 condor off -master
#3 log in as root, then start the condor master using the path from #1

This will take care of one machine...

Hope this helps, 

Mike Yoder
Principal Member of Technical Staff
Direct : +1.408.321.9000
Fax    : +1.408.904.5992
Mobile : +1.408.497.7597
yoderm@xxxxxxxxxx

Optena Corporation
2860 Zanker Road, Suite 201
San Jose, CA 95134
http://www.optena.com