kill: send signals to processes

Updated November 2026

kill sends a signal to a process. Most signals ask the process to do something, usually to stop. Despite the name, killing is only one of the things it can do.

kill [-SIGNAL] PID ... pkill [-SIGNAL] [options] pattern killall [-SIGNAL] name # Linux

The signals you will use

Signal	Number	Meaning
`TERM`	15	Please stop. The default; the program can clean up first.
`INT`	2	Interrupt, the same as `Ctrl`+`C`.
`HUP`	1	Hang-up. Many daemons reload their configuration on HUP.
`KILL`	9	Stop immediately. Cannot be caught or ignored; no cleanup.
`STOP` / `CONT`	19 / 18 on Linux	Pause and resume a process.
`USR1`, `USR2`	10, 12 on Linux	Program-defined, for example "reopen log files".

Numbers differ between systems for some signals; names work everywhere. kill -l lists them.

Examples

kill 4122                     # ask process 4122 to stop (TERM)
kill -HUP 1022                # ask nginx's master process to reload
kill -9 4122                  # force, if it ignored TERM
kill -0 4122 && echo alive     # does the process exist (and may I signal it)?
pkill -f "python3 report.py"  # by command line
pkill -u anna                 # every process of a user
killall firefox               # by name (Linux)
pgrep -a java                 # check what a pattern matches before using pkill

killall is different on Solaris and AIX

On System V Unix, killall is a shutdown command that kills every process. Use pkill on Solaris; on AIX use kill with PIDs from ps -ef.

Why kill -9 is the last resort

KILL stops a program in the middle of whatever it was doing: it cannot finish writing files, release locks, remove temporary files or tell its child processes. Databases may need recovery afterwards. Send TERM first, wait a few seconds, and only then use -9. A process in state D (see ps) may not die even from KILL until its disk or network operation finishes.

Services

For services started by systemd, use systemctl stop, restart or reload instead of kill, so systemd knows what happened and does not restart the service unexpectedly. See systemctl. To send a signal through systemd: systemctl kill -s HUP nginx.

Jobs in your shell

long-command &          # run in the background
jobs                    # list background jobs of this shell
fg %1                   # bring job 1 to the foreground
# Ctrl+Z pauses the foreground job, then:
bg                      # continue it in the background
kill %1                 # signal a job by its job number
nohup ./backup.sh &     # keep running after you log out (output in nohup.out)
disown %1               # remove a running job from the shell, so logging out does not stop it

For long work on remote servers, a terminal multiplexer is more convenient than nohup: see tmux.