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

Re: [Condor-users] condor_submit hangs when Queue > 1



On Thu, Dec 30, 2010 at 09:09:09AM -0500, Michael Hanke wrote:
> Therefore the problem is either in the way I build Condor or in the
> substantial diff between
> 
> 0dbdaa1bc6319a6cde6ad0c358e0673af43e915b
> 
> and 
> 
> cc9972363cdaa7d262ca1003765d81747ed69d73
> 
> (both touching schedd and condor_submit code).

I made some progress. First I updated the code to today's content of the
master branch, but that didn't change the behaviour. Next I removed all
patches that I developed to make the Debian package compliant with
Debian policy and the build environment of Debian squeeze. I could
narrow it down to the fact that I have built condor's internal libs as
shared libraries. Once I removed this patch I still couldn't submit, but
I got an error message that wasn't there before:

michael@meiner /tmp % condor_submit job
Submitting job(s)
[ some delay but much shorter than before ]
ERROR: Failed to create cluster
1 michael@meiner /tmp % tail /var/log/condor/SchedLog
12/30/10 10:55:12 (pid:15619)    /etc/condor/condor_config.local
12/30/10 10:55:12 (pid:15619) DaemonCore: command socket at <192.168.0.13:52620>
12/30/10 10:55:12 (pid:15619) DaemonCore: private command socket at <192.168.0.13:52620>
12/30/10 10:55:12 (pid:15619) Setting maximum accepts per cycle 4.
12/30/10 10:55:12 (pid:15619) History file rotation is enabled.
12/30/10 10:55:12 (pid:15619)   Maximum history file size is: 20971520 bytes
12/30/10 10:55:12 (pid:15619)   Number of rotated history files is: 2
12/30/10 10:55:12 (pid:15619) "/usr/sbin/condor_shadow.std -classad" did not produce any output, ignoring
12/30/10 10:55:40 (pid:15619) PERMISSION DENIED to michael@meiner from host 192.168.0.13 for command 1112 (QMGMT_WRITE_CMD), access level WRITE: reason: WRITE authorization policy contains no matching ALLOW entry for this request; identifiers used for this host: 192.168.0.13,192.168.0.13

Apparently

  ALLOW_WRITE = $(FULL_HOSTNAME)

is not sufficient for a personal condor installation. After I changed
that to

  ALLOW_WRITE = *

I can

michael@meiner /tmp % condor_submit job
Submitting job(s)..
2 job(s) submitted to cluster 1.

Of course this is not a sensible setting and $(FULL_HOSTNAME) should do
it, right?


Michael

-- 
Michael Hanke
http://mih.voxindeserto.de