CONDOR_OFF(1)

📦
version 25.14.1 · 12 versions
, in fedora 43

👑 condor_off — the HTCondor Manual, but make it cute

okay bestie, gather round, because we're talking about condor_off from the HTCondor manual and honestly? it's giving "politely ask your daemons to go night-night" energy. 💤✨

✨ NAME ✨

condor_off — Shutdown HTCondor daemons 🫶

🌸 SYNOPSIS 🌸

condor_off [-help | -version ]

condor_off [-graceful | -fast | -peaceful | -force-graceful | -drain ]
[-annex name] [-debug[:opts] ] [-pool
centralmanagerhostname[:portnumber]] [ -name hostname | hostname | -addr
"<a.b.c.d:port>" | "<a.b.c.d:port>" | -constraint expression | -all ]
[-daemon daemonname | -master] [-exec name] [-reason "reason-string"]
[-request-id id] [-check expr] [-start expr]

💅 DESCRIPTION 💅

So like, condor_off shuts down a set of the HTCondor daemons running on a set of one or more machines. Ngl, by default it does this cleanly, so that jobs actually get time to shut down instead of being yanked mid-sentence. 😭

The command condor_off with no arguments at all will shut down all daemons except condor_master — unless -annex name is specified, which changes the whole vibe. And why do we spare condor_master? Because it can then handle both local and remote requests to restart the other HTCondor daemons if need be. It's the main character fr. 🌟 To restart HTCondor running on a machine, go see the condor_on command.

When you choose the -drain option, draining options can be specified using the optional -reason, -request-id, -check, and -start arguments. Slay, options within options. 🎀

With the -daemon master option, condor_off will shut down all daemons including the condor_master — no survivors, periodt. Specification using the -daemon option will shut down only the specified daemon. Pick your fighter. 💀

When shutting down all daemons including the condor_master, the -exec argument can be used to tell the master to run a configured MASTER_SHUTDOWN_<Name> script before it exits. Iconic exit ritual honestly. 🕯️

For security reasons of authentication and authorization, this command requires ADMINISTRATOR level of access. No random people off the street, bestie, this is a locked-door situation. 👑

🦋 OPTIONS 🦋

  • -help — Display usage information. The "what does this even do" button. 🙈
  • -version — Display version information. Receipts of what you're running. 🧾
  • -graceful — The default. If jobs are running, wait for up to the configured grace period for them to finish, then exit. Chill but not too chill. ⏳
  • -fast — Quickly shutdown daemons, immediately evicting any running jobs. A minimum of the first two characters of this option must be specified, to distinguish it from the -force-graceful command. It's giving speedrun energy, no goodbyes. 💨
  • -peaceful — Wait indefinitely for jobs to finish. Not a single deadline in sight, we simply wait. 🧘‍♀️
  • -force-graceful — Force a graceful shutdown, even after issuing a -peaceful command. A minimum of the first two characters of this option must be specified, to distinguish it from the -fast command. It's the "okay actually change of plans" option. 🙈
  • -drain — Send a condor_drain command with the -exit-on-completion option to all condor_startd daemons that are managed by this master. Then wait for all condor_startd daemons to exit before shutting down other daemons. Very "we leave in an orderly fashion" of it. 🚪
  • -reason "reason-string" — Use with -drain to set a -reason "reason-string" value for the condor_drain command. Tell them why bestie, closure matters. 💌
  • -request-id id — Use with -drain to set a -request-id id value for the condor_drain command. Gotta have that ID, iykyk. 🔖
  • -check expr — Use with -drain to set a -check expr value for the condor_drain command. 🧐
  • -start expr — Use with -drain to set a -start expr value for the condor_drain command. 🚀
  • -annex name — Turn off master daemons in the specified annex. By default this will result in the corresponding instances shutting down. Bye annex, love u, mean it. 👋
  • -debug[:opts] — Causes debugging information to be sent to stderr. The debug level can be set by specifying an optional opts value. Otherwise, the configuration variable TOOL_DEBUG sets the debug level. Spill the tea (to stderr). 🍵
  • -pool centralmanagerhostname[:portnumber] — Specify a pool by giving the central manager's host name and an optional port number. Aim at a whole different pool, lowkey powerful. 🏊
  • -name hostname — Send the command to a machine identified by hostname. 📮
  • hostname — Send the command to a machine identified by hostname. (Yeah, you can just say the name, no flag needed, we love an effortless queen.) ✨
  • -addr "<a.b.c.d:port>" — Send the command to a machine's master located at "<a.b.c.d:port>". IP-address-coded. 🔢
  • "<a.b.c.d:port>" — Send the command to a machine located at "<a.b.c.d:port>". 🎯
  • -constraint expression — Apply this command only to machines matching the given ClassAd expression. Filtered, curated, algorithm-approved. 🪄
  • -all — Send the command to all machines in the pool. Everyone. All of them. The whole group chat. 📢
  • -master — Shutdown the condor_master after shutting down all other daemons. The final boss exits last. 👑
  • -exec name — When used with -master, the condor_master will run the program configured as MASTER_SHUTDOWN_<Name> after shutting down all other daemons. 🏃‍♀️
  • -daemon daemonname — Send the command to the named daemon. Without this option, the command is sent to the condor_master daemon. Choose your target bestie. 🎪

🌈 GRACEFUL VS. PEACEFUL VS FAST 🌈

A "fast" shutdown will cause the requested daemon to exit. Jobs running under a startd that is shutdown fast will be evicted. Jobs running on a schedd that is shutdown fast will be left running for their job lease duration (default of 20 minutes). (That is, assuming the corresponding startd is not also being shut down). If that schedd restarts before the job lease expires, it will reconnect to these running jobs and continue to run them, as long as the schedd and startd are running. It's the "leaving but my location is still on" situation fr. 📍

A "graceful" shutdown of a schedd is functionally the same as a "fast" shutdown of a schedd. Same energy, different name, no cap. 🤝

A "graceful" shutdown of a startd that has jobs running under it causes the startd to wait for the jobs to exit of their own accord, up to the MaxJobRetirementTime. After the MaxJobRetirementTime, the startd will evict any remaining running jobs and exit. Respectful, but with a hard deadline. ⏰

A "peaceful" shutdown of a startd or schedd will cause that daemon to wait indefinitely for all existing jobs to exit before shutting down. During this time, no new jobs will start. It will literally wait forever, bestie, that's the whole personality. 🕰️

🍓 EXIT STATUS 🍓

condor_off will exit with a status value of 0 (zero) upon success, and it will exit with the value 1 (one) upon failure. Zero = slay, one = we do not talk about it. 💀

🧚 EXAMPLES 🧚

To shut down all daemons (other than condor_master) on the local host:

$ condor_off

To shut down only the condor_collector on three named machines:

$ condor_off  cinnamon cloves vanilla -daemon collector

To shut down daemons within a pool of machines other than the local pool, use the -pool option. The argument is the name of the central manager for the pool. Note that one or more machines within the pool must be specified as the targets for the command. This command shuts down all daemons except the condor_master on the single machine named cae17 within the pool of machines that has condor.cae.wisc.edu as its central manager:

$ condor_off  -pool condor.cae.wisc.edu -name cae17

💖 AUTHOR 💖

HTCondor Team — the collective group-chat admins of HTCondor, we stan. 🫶

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!! 🎀 it came from the HTCondor Manual, dated Oct 16, 2026, and it's filed under CONDOR_OFF(1). go shut down some daemons responsibly, ily, bye!! 💅✨🦋

📦 where this page comes from