[Condor-users] 12 second delay when submitting a DAG

[David Arthur <mumrah@xxxxxxxxx>]

When submitting a DAG, I see the following output:

10/05 13:29:23 ******************************************************
10/05 13:29:23 ** condor_scheduniv_exec.805.0 (CONDOR_DAGMAN) STARTING UP
10/05 13:29:23 ** /usr/bin/condor_dagman
10/05 13:29:23 ** SubsystemInfo: name=DAGMAN type=DAGMAN(10) class=DAEMON(1)
10/05 13:29:23 ** Configuration: subsystem:DAGMAN local:<NONE> class:DAEMON
10/05 13:29:23 ** $CondorVersion: 7.4.3 Aug  4 2010 BuildID: 261829 $
10/05 13:29:23 ** $CondorPlatform: X86_64-LINUX_RHEL5 $
10/05 13:29:23 ** PID = 20233
10/05 13:29:23 ** Log last touched time unavailable (No such file or directory)
10/05 13:29:23 ******************************************************
10/05 13:29:23 Using config source: /etc/condor/condor_config
10/05 13:29:23 Using local config sources:
10/05 13:29:23    /etc/condor/condor_config.local
10/05 13:29:23 DaemonCore: Command Socket at <192.168.10.58:34064>
10/05 13:29:23 DAGMAN_DEBUG_CACHE_SIZE setting: 5242880
10/05 13:29:23 DAGMAN_DEBUG_CACHE_ENABLE setting: False
10/05 13:29:23 DAGMAN_SUBMIT_DELAY setting: 0
10/05 13:29:23 DAGMAN_MAX_SUBMIT_ATTEMPTS setting: 6
10/05 13:29:23 DAGMAN_STARTUP_CYCLE_DETECT setting: 0
10/05 13:29:23 DAGMAN_MAX_SUBMITS_PER_INTERVAL setting: 5
10/05 13:29:23 DAGMAN_USER_LOG_SCAN_INTERVAL setting: 2
10/05 13:29:23 allow_events (DAGMAN_IGNORE_DUPLICATE_JOB_EXECUTION,
DAGMAN_ALLOW_EVENTS) setting: 114
10/05 13:29:23 DAGMAN_RETRY_SUBMIT_FIRST setting: 1
10/05 13:29:23 DAGMAN_RETRY_NODE_FIRST setting: 0
10/05 13:29:23 DAGMAN_MAX_JOBS_IDLE setting: 0
10/05 13:29:23 DAGMAN_MAX_JOBS_SUBMITTED setting: 0
10/05 13:29:23 DAGMAN_MUNGE_NODE_NAMES setting: 1
10/05 13:29:23 DAGMAN_PROHIBIT_MULTI_JOBS setting: 0
10/05 13:29:23 DAGMAN_SUBMIT_DEPTH_FIRST setting: 0
10/05 13:29:23 DAGMAN_ABORT_DUPLICATES setting: 1
10/05 13:29:23 DAGMAN_ABORT_ON_SCARY_SUBMIT setting: 1
10/05 13:29:23 DAGMAN_PENDING_REPORT_INTERVAL setting: 600
10/05 13:29:23 DAGMAN_AUTO_RESCUE setting: 1
10/05 13:29:23 DAGMAN_MAX_RESCUE_NUM setting: 100
10/05 13:29:23 DAGMAN_DEFAULT_NODE_LOG setting: null
10/05 13:29:23 ALL_DEBUG setting:
10/05 13:29:23 DAGMAN_DEBUG setting:
10/05 13:29:23 argv[0] == "condor_scheduniv_exec.805.0"
10/05 13:29:23 argv[1] == "-Debug"
10/05 13:29:23 argv[2] == "3"
10/05 13:29:23 argv[3] == "-Lockfile"
10/05 13:29:23 argv[4] == "submit.dag.lock"
10/05 13:29:23 argv[5] == "-AutoRescue"
10/05 13:29:23 argv[6] == "1"
10/05 13:29:23 argv[7] == "-DoRescueFrom"
10/05 13:29:23 argv[8] == "0"
10/05 13:29:23 argv[9] == "-Dag"
10/05 13:29:23 argv[10] == "submit.dag"
10/05 13:29:23 argv[11] == "-CsdVersion"
10/05 13:29:23 argv[12] == "$CondorVersion: 7.4.3 Aug  4 2010 BuildID: 261829 $"
10/05 13:29:23 Default node log file is:
</var/lib/condor/spool/cluster805.proc0.subproc0/submit.dag.nodes.log>
10/05 13:29:23 DAG Lockfile will be written to submit.dag.lock
10/05 13:29:23 DAG Input file is submit.dag
10/05 13:29:23 Parsing 1 dagfiles
10/05 13:29:23 Parsing submit.dag ...
10/05 13:29:23 Dag contains 1 total jobs
10/05 13:29:23 Sleeping for 12 seconds to ensure ProcessId uniqueness


I would really like to avoid the 12 second sleep as this is a
low-latency application. Is there a configuration or argument for
`condor_dagman` that will avoid this sleep?

Cheers,

David Arthur