CONDOR_VACATE_JOB(1)

๐Ÿ’… condor_vacate_job โ€” bestie's guide to HTCondor

okay so real talk, this is a page from the HTCondor Manual, and it's all about condor_vacate_job. ๐Ÿฆ‹


โœจ NAME โœจ

condor_vacate_job โ€” the HTCondor Manual's own little diva of a command. She stops running job(s) and hands the resources back to the host machine(s) that were executing them. it's giving "you're done here, thank u next" ๐Ÿ’–


โœจ SYNOPSIS โœจ

condor_vacate_job [-help | -version]

condor_vacate_job [OPTIONS] [cluster... | cluster.proc... | user...]

condor_vacate_job [-debug] [-long] [-totals] [-all] [-constraint
expression] [-pool hostname[:portnumber] | -name scheddname | -addr
"<a.b.c.d:port>"] [-fast]

no cap, that's literally the whole shape of it ๐Ÿ’ซ


โœจ DESCRIPTION โœจ

So what she does: she vacates job(s) off the host machine(s) where they're currently running ๐Ÿƒโ™€๏ธ. Resources go back to the host machine(s), and the job(s) go right back to idle so that resource matchmaking and execution can happen somewhere else. iconic behavior honestly.

And here's the cute part โ€” vacating the job makes the condor accountant stop charging the user in terms of priority. your budget said "not today" ๐Ÿ’ธโœจ

But listen ๐Ÿ‘€ for any given job, only these people can vacate it:

  • the owner of the job, or
  • one of the queue super users โ€” defined by the QUEUE_SUPER_USERS macro.

periodt. that's the whole guest list ๐ŸŽ€


โœจ OPTIONS โœจ

  • -help โ€” Display usage information. a little "how do u do, fellow kids" moment ๐Ÿซถ
  • -version โ€” Display version information. she'll tell u who she is ๐Ÿ’โ™€๏ธ
  • -long โ€” Display result ClassAd. the full lore ๐Ÿ“œ
  • -totals โ€” Display success/failure totals. receipts! ๐Ÿงพ
  • -fast โ€” Perform a fast vacate and hard kill the jobs. no soft launch, we're going straight to the hard kill ๐Ÿ’€๐Ÿ”ฅ
  • -pool hostname[:portnumber]` โ€” Specify a pool by giving the central manager's host name and an optional port number. โœจ
  • -name scheddname` โ€” Send the command to a machine identified by scheddname. ๐Ÿ“ฎ
  • -addr "<a.b.c.d:port>" โ€” Send the command to a machine located at "<a.b.c.d:port>". straight to the address, iykyk ๐Ÿ 
  • -constraint expression` โ€” Vacate all jobs which match the job ClassAd expression constraint. filtering? we love to see it ๐ŸŽฏ
  • -all โ€” Vacate all the jobs in the queue. everyone. all of them. the whole group chat ๐Ÿ’…
  • cluster โ€” Vacate all jobs in the specified cluster. the whole clique ๐ŸŽ€
  • cluster.process โ€” Vacate the specific job in the cluster. just the one bestie ๐Ÿฆ‹
  • user โ€” Vacate jobs belonging to specified user. it's giving targeted ๐Ÿ‘‘

๐ŸŒธ GENERAL REMARKS ๐ŸŒธ

okay important stuff, listen up bestie ๐Ÿซถ

If the -name option is specified, the named condor_schedd is targeted for processing. Otherwise, the local condor_schedd is targeted. lowkey simple.

Vacated jobs will be sent a soft kill signal (SIGTERM by default, or the value of KillSig in the job's ClassAd) โ€” unless -fast is used, in which case, yeah, hard kill, no softness ๐Ÿ˜ญ

Using condor_vacate_job on jobs which are not currently running has no effect. nothing. zero. the silence ๐Ÿ’€

Jobs running in the local or scheduler universe can be vacated. And since these types of jobs run directly on the Access Point, they will always restart on the same resources and may begin executing again immediately if there is a low load or number of jobs submitted to the Access Point. so likeโ€ฆ she might just come right back ๐Ÿ™ˆ

DAGMan can also be vacated, and it will enter recovery mode upon restarting. BUT โ€” and this is the big but โ€” all jobs submitted and managed by DAGMan will continue running uninterrupted. ๐Ÿ”ฅ

WARNING: Do not confuse this tool with condor_vacate. condor_vacate is intended for use by the owners/administrators of an Execution Point to evict jobs from their host machine. different girl, different job, do not mix them up bestie ๐Ÿ˜ค

โœจ EXIT STATUS โœจ

  • 0 โ€” Success ๐ŸŽ‰
  • 1 โ€” Failure has occurred ๐Ÿ˜ญ

๐Ÿ’– EXAMPLES ๐Ÿ’–

To vacate a specific job:

$ condor_vacate_job 23.0

To vacate a job fast:

$ condor_vacate_job -fast 23.0

To vacate all jobs owned by user Mary:

# condor_vacate_job mary

To vacate all vanilla universe jobs owned by Mary:

# condor_vacate_job -constraint 'JobUniverse == 5 && Owner == "mary"'

look at her go ๐ŸŒˆ


๐Ÿฆ‹ SEE ALSO ๐Ÿฆ‹

condor_vacate, condor_rm, condor_continue, condor_suspend, condor_hold, condor_release

the whole friend group honestly ๐Ÿซถ


โœจ AVAILABILITY โœจ

Linux, MacOS, Windows โ€” she's everywhere, multi-platform queen ๐Ÿ‘‘


๐ŸŒˆ AUTHOR ๐ŸŒˆ

HTCondor Team ๐Ÿ’œ


1990-2026, Center for High Throughput Computing, Computer Sciences Department, University of Wisconsin-Madison, Madison, WI, US. Licensed under the Apache License, Version 2.0.


and that's the whole page bestie โœจ this was CONDOR_VACATE_JOB(1) from the HTCondor Manual, dated Oct 16, 2026 โ€” go forth, vacate responsibly, and remember: never confuse her with condor_vacate ๐Ÿ’…๐Ÿซถ๐Ÿ’–

๐Ÿ“ฆ where this page comes from