I can't get condor_submit -addr to work when condor_schedd is behind a condor_shared_port.
Output from condor_submit is below
sh-4.2$ condor_submit -debug -addr "<126.96.36.199:9618>" job.sub
Submitting job(s)07/04/18 10:06:49 condor_read() failed: recv(fd=4) returned -1, errno = 104 Connection reset by peer, reading 5 bytes from schedd at <188.8.131.52:9618>.
07/04/18 10:06:49 IO: Failed to read packet header
07/04/18 10:06:49 SECMAN: no classad from server, failing
ERROR: Failed to connect to local queue manager
SECMAN:2007:Failed to end classad message.
Error message written to /var/log/condor/SharedPortLog in the schedd container is
07/04/18 10:06:49 SharedPortServer: server was busy, failed to connect collector as requested by <184.108.40.206:46528>: primary (7d2cc1f5fc7f6a4e2eb39facb9bb27877fdd809e4b7fa28fd830cd99c77172ee/collector): Connection refused (111); alt (/var/lock/condor/daemon_sock/collector): Connection refused (111)
Nothing is written to /var/log/condor/SchedLog
Why is condor_submit even trying to access the collector when -addr is meant to tell it to connect straight to the sched? Is there is a bug in condor_submit that means it asks the shared_port_daemon to connect to the the collector, not the sched, even when the -addr option it set?
Everything works fine when sched isn't running behind a condor_shared_port, so I've worked round this issue by simply not using a shared port.
Relevant versions are
$CondorVersion: 8.6.11 May 10 2018 BuildID: 440910 $
$CondorPlatform: x86_64_RedHat7 $
Relevant files are
sh-4.2$ cat job.sub
should_transfer_files = YES
when_to_transfer_output = ON_EXIT
Universe = vanilla
Executable = /bin/bash
Arguments = test.sh
Log = job.log
Output = job.out
Error = job.error
transfer_input_files = test.sh
sh-4.2$ cat test.sh
echo Starting test.sh
echo Ending test.sh
I'm running HTCondor in a container on Kubernetes, but doubt that is relevant to this problem.