CONDOR_TOP(1)

📦
version 25.14.1 · 12 versions
, in fedora 43

🌸 NAME 🌸

condor_top — it's in the HTCondor Manual, bestie 💖

Displays the status and runtime statistics of a HTCondor daemon. She's giving ✨ dashboard energy ✨ and she ate.

✨ SYNOPSIS ✨

condor_top [-h ]

condor_top [-l ] [-H ] [-p centralmanagerhostname[:portname]] [-n name]
[-d delay] [-c columnset] [-s sortcolumn] [--attrs=<attr1,attr2,...>]
[daemon options ]

condor_top [-c columnset] [-s sortcolumn] [--attrs=<attr1,attr2,...>]
[classad-filename classad-filename ]

💅 DESCRIPTION 💅

okay so condor_top shows you the status of a HTCondor daemon — like, memory usage, duty cycle, all of it — and then it calculates and displays runtime statistics for that daemon's subprocesses. She's doing the math so you don't have to. 👑

When you give her no arguments at all, condor_top shows the status for the primary daemon, picked by looking at what role the current machine has — she does this by scanning the DAEMON_LIST configuration setting. And if multiple daemons are listed, she monitors one of these, in decreasing priority:

  1. condor_schedd
  2. condor_startd
  3. condor_collector
  4. condor_negotiator
  5. condor_master

If the condor_collector hands back multiple ClassAds for the daemon type you picked, condor_top will display stats from the first ClassAd returned. (First come first served, no cap 💀) You can narrow it down by passing the NAME of a specific daemon with -n.

The default delay is STATISTICS_WINDOW_QUANTUM, which is 4 minutes (240 seconds) in a default HTCondor configuration. Setting the delay smaller can genuinely help you catch spikes of activity 🦋 — but setting it too small will lead to poor measurements of the duty cycle and of the runtime statistics. So, like, don't be messy about it.

condor_top can run in a top-like "live" mode by passing -l. Live mode is similar to the *nix top command, with stats updating every delay seconds. Redirecting stdout will disable live mode even if -l is set. To exit condor_top while in live mode, hit Ctrl+C. 😭

Lowkey important: if you -l and pipe it somewhere, live mode is cancelled. The shell said "not today."

condor_top can also be passed two files containing ClassAds from the same HTCondor daemon. In that case the condor_collector will not be queried — instead the statistics are computed and displayed immediately from the two ClassAds. Only -c, -s, and -attrs options are considered when passing ClassAds via files. 🫶

When monitoring a condor_schedd, passing -H (or -history) reads the two most recent daemon ClassAds from the $(SPOOL)/schedd_daemon_history file instead of querying the schedd twice. This mode disables live output, cannot be combined with file input, and ignores the -p option because the data are sourced locally. iykyk 💖

🎀 The Subprocess Stat Columns 🎀

The following subprocess stat columns may be displayed (* = default, queen behavior 👑). Each column is printed with a three-row header followed by a horizontal rule separating the header from the data.

  • Row 1 names the time scope: Recent = between the two ClassAds, Lifetime/Total = since daemon start
  • Row 2 names the metric
  • Row 3 gives the units: s = seconds, # = count, /s = count per second, % = percentage

The displayed header is shown in quotes below; the identifier following it is the name to pass to the -s option to sort by that column. 💅

  • "Name" (Item) — *Name of the subprocess
  • "Recent Runtime" (InstRt) — *Total runtime between the two ClassAds
  • "Recent Avg" (InstAvg) — *Mean runtime per execution between the two ClassAds
  • "Total Runtime" (TotalRt) — Total runtime since daemon start
  • "Lifetime Avg" (TotAvg) — *Mean runtime per execution since daemon start
  • "Lifetime Max" (TotMax) — *Max runtime per execution since daemon start
  • "Lifetime Min" (TotMin) — Min runtime per execution since daemon start
  • "Pct of Avg" (RtPctAvg) — *Percent of mean runtime per execution. The ratio of InstAvg to TotAvg, expressed as a percentage
  • "Pct of Max" (RtPctMax) — Percent of max runtime per execution. The ratio of (InstAvg - TotMin) to (TotMax - TotMin), expressed as a percentage
  • "Std Devs" (RtSigmas) — Standard deviations from mean runtime. The ratio of (InstAvg - TotAvg) to the standard deviation in runtime per execution since daemon start
  • "Recent Count" (InstCt) — Executions between the two ClassAds
  • "Recent Rate" (InstRate) — *Executions per second between the two ClassAds
  • "Total Count" (TotalCt) — Total executions (counts) since daemon start
  • "Lifetime Rate" (AvgRate) — *Mean count rate. Executions per second since daemon start
  • "Pct of Rate" (CtPctAvg) — Percent of mean count rate. The ratio of InstRate to AvgRate, expressed as a percentage.

🌈 OPTIONS 🌈

  • -h — Displays the list of options. The classic "help me bestie" move.
  • -l — Puts condor_top into a live, continually updating mode. She's always watching. 👀
  • -H, -history, --history — When monitoring a condor_schedd, read the two most recent daemon ClassAds from $(SPOOL)/schedd_daemon_history instead of querying the daemon twice. This mode disables live updates, ignores -p, and cannot be combined with file input.
  • -p centralmanagerhostname[:portname] — Query the daemon via the specified central manager. If omitted, the value of the configuration variable COLLECTOR_HOST is used.
  • -n name — Query the daemon named name. If omitted, the value used will depend on the type of daemon queried (see Daemon Options). 🦋
  • -d delay — Specifies the delay between ClassAd updates, in integer seconds. If omitted, the value of the configuration variable STATISTICS_WINDOW_QUANTUM is used.
  • -c columnset — Display columnset set of columns. Valid *columnset*s are: default, runtime, count, all.
  • -s sortcolumn — Sort table by sortcolumn. Defaults to InstRt. Sorting is the moment. ⭐
  • --attrs=<attr1,attr2,...> — Comma-delimited list of additional ClassAd attributes to monitor.

👑 Daemon Options 👑

  • -collector — Monitor condor_collector ClassAds. If -n is not set, the constraint "Machine == COLLECTOR_HOST" will be used.
  • -negotiator — Monitor condor_negotiator ClassAds. If -n is not set, the constraint "Machine == COLLECTOR_HOST" will be used.
  • -master — Monitor condor_master ClassAds. If -n is not set, the constraint "Machine == COLLECTOR_HOST" will be used.
  • -schedd — Monitor condor_schedd ClassAds. If -n is not set, the constraint "Machine == FULL_HOSTNAME" will be tried, otherwise the first condor_schedd ClassAd returned from the condor_collector will be used.
  • -startd — Monitor condor_startd ClassAds. If -n is not set, the constraint "Machine == FULL_HOSTNAME" will be tried, otherwise the first condor_startd ClassAd returned from the condor_collector will be used.

🌸 AUTHOR 🌸

HTCondor Team 🫶 — built different, fr fr.

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.

okay that's a wrap bestie 💖 this was CONDOR_TOP(1) from the HTCondor Manual, dated Oct 16, 2026. go monitor something iconic, mwah 💅✨🎀

📦 where this page comes from