The Wiki has moved to new hardware, and the old server died.

Welcome to the Slackware Documentation Project

This is an old revision of the document!


Process Control

Slackware systems often run hundreds or thousands of programs, each of which is referred to as a process. Managing these processes is an important part of system administration. So how exactly do we handle all of these seperate processes?

ps

The first step in managing processes is figuring out what processes are currently running. The most popular and powerful tool for this is ps(1). Without any arguments, ps won't tell you much information. By default, it only tells you what processes are running in your currently active shell. If we want more information, we'll need to look deeper.

darkstar:~$ ps
  PID TTY          TIME CMD
12220 pts/4    00:00:00 bash
12236 pts/4    00:00:00 ps

Here you can see what processes you are running in your currently active shell or terminal and only some information is included. The PID is the “Process ID”; every process is assigned a unique number. The TTY tells you what terminal device the process is attached to. Naturally, CMD is the command that was run. You might be a little confused by TIME though, since it seems to move so slowly. This isn't the amount of real time the process has been running, but rather the amount of CPU time the process has consumed. An idle process uses virtually no CPU time, so this value may not increase quickly.

Viewing only our own processes isn't very much fun, so let's take a look at all the processes on the system with the [-e] argument.

darkstar:~$ ps -e
  PID TTY          TIME CMD
    1 ?        00:00:00 init
    2 ?        00:00:00 kthreadd
    3 ?        00:00:00 migration/0
    4 ?        00:00:00 ksoftirqd/0
    7 ?        00:00:11 events/0
    9 ?        00:00:01 work_on_cpu/0
   11 ?        00:00:00 khelper
  102 ?        00:00:02 kblockd/0
  105 ?        00:01:19 kacpid
  106 ?        00:00:01 kacpi_notify
... many more lines omitted ...

The above example uses the standard ps syntax, but much more information can be discovered if we use BSD syntax. In order to do so, we must use the [aux] argument.

This is distinct from the [-aux] argument, but in most cases the two arguments are equivilant. This is a decades-old relic. For more information, see the man page for ps.
darkstar:~$ ps aux
USER       PID %CPU %MEM    VSZ   RSS TTY      STAT START   TIME COMMAND
root         1  0.0  0.0   3928   632 ?        Ss   Apr05   0:00 init [3]  
root         2  0.0  0.0      0     0 ?        S    Apr05   0:00 [kthreadd]
root         3  0.0  0.0      0     0 ?        S    Apr05   0:00 [migration/0]
root         4  0.0  0.0      0     0 ?        S    Apr05   0:00 [ksoftirqd/0]
root         7  0.0  0.0      0     0 ?        S    Apr05   0:11 [events/0]
root         9  0.0  0.0      0     0 ?        S    Apr05   0:01 [work_on_cpu/0]
root        11  0.0  0.0      0     0 ?        S    Apr05   0:00 [khelper]
... many more lines omitted ....

As you can see, BSD syntax offers much more information, including what user controls the process and what percentage of RAM and CPU the process is consuming when ps is run.

To accomplish bits of this, on a per process basis, ps allows one or more process IDs (PIDs) to be provided in the command line, and has the '-o' flag to show a particular attribute of the PID.

darkstar:~$ ps -o cmd -o etime $$
CMD                             ELAPSED
/bin/bash                         12:22

What this is displaying, is the PID's command name (cmd), and its elapsed time (etime). The PID in this example, is a shell variable for the PID of the current shell. So you can see, in this example, the shell process has existed for 12 minutes, 22 seconds.

Using the pgrep(1) command, this can get more automatable.

darkstar:~$ ps -o cmd -o rss -o vsz $(pgrep httpd)
CMD                           RSS    VSZ
/usr/sbin/httpd -k restart  33456  84816
/usr/sbin/httpd -k restart  33460  84716
/usr/sbin/httpd -k restart  33588  84472
/usr/sbin/httpd -k restart  30424  81608
/usr/sbin/httpd -k restart  33104  84900
/usr/sbin/httpd -k restart  33268  85112
/usr/sbin/httpd -k restart  30640  82724
/usr/sbin/httpd -k restart  15168  67396
/usr/sbin/httpd -k restart  33180  84416
/usr/sbin/httpd -k restart  33396  84592
/usr/sbin/httpd -k restart  32804  84232

In this example, a subshell execution, using pgrep, is returning the PIDs of any process, whose command name includes 'httpd'. Then ps displaying the command name, resident memory size, and virtual memory size.

Finally, ps can also create a process tree. This shows you which processes have children processes. Ending the parent of a child process also ends the child. We do this with the [-ejH] argument.

darkstar:~$ ps -ejH
... many lines omitted ...
 3660  3660  3660 tty1     00:00:00   bash
29947 29947  3660 tty1     00:00:00     startx
29963 29947  3660 tty1     00:00:00       xinit
29964 29964 29964 tty7     00:27:11         X
29972 29972  3660 tty1     00:00:00         sh
29977 29972  3660 tty1     00:00:05           xscreensaver
29988 29972  3660 tty1     00:00:04           xfce4-session
29997 29972  3660 tty1     00:00:16             xfwm4
29999 29972  3660 tty1     00:00:02             Thunar
... many more lines omitted ...

As you can see, ps(1) is an incredibly powerful tool for determining not only what processes are currently active on your system, but also for learning lots of important information about them.

As is the case with many of the applications, there is often several tools for the job. Similar to the ps -ejH output, but more terse, is pstree(1). It displays the process tree, a bit more visually.

darkstar:~$ pstree
init-+-atd
     |-crond
     |-dbus-daemon
     |-httpd---10*[httpd]
     |-inetd
     |-klogd
     |-mysqld_safe---mysqld---8*[{mysqld}]
     |-screen-+-4*[bash]
     |        |-bash---pstree
     |        |-2*[bash---ssh]
     |        `-bash---irssi
     |-2*[sendmail]
     |-ssh-agent
     |-sshd---sshd---sshd---bash---screen
     `-syslogd

kill and killall

Managing processes isn't only about knowing which ones are running, but also about communicating with them to change their behavior. The most common way of managing a program is to terminate it. Thus, the tool for the job is named kill(1). Despite the name, kill doesn't actually terminate processes, but sends signals to them. The most common signal is a SIGTERM, which tells the process to finish up what it is doing and terminate. There are a variety of other signals that can be sent, but the three most common are SIGTERM, SIGHUP, and SIGKILL.

What a process does when it receives a signal varies. Most programs will terminate (or attempt to terminate) whenever they receive any signal, but there are a few important differences. For starters, the SIGTERM signal informs the process that it should terminate itself at its earliest convenience. This gives the process time to finish up any important activities, such as writing information to the disk, before it closes. In contrast, the SIGKILL signal tells the process to terminate itself immediately, no questions asked. This is most useful for killing processes that are not responding and is sometimes called the “silver bullet”. Some processes (particularly daemons) capture the SIGHUP signal and reload their configuration files whenever they receive it.

In order to signal a process, we first need to know it's PID. You can get this easily with ps as we discused. In order to send different signals to a running process, you simply pass the signal number and [-s] as an argument. The [-l] argument lists all the signals you can choose and their number. You can also send signals by their name with [-s].

darkstar:~$ kill -l
 1) SIGHUP	 2) SIGINT	 3) SIGQUIT	 4) SIGILL
 5) SIGTRAP	 6) SIGABRT	 7) SIGBUS	 8) SIGFPE
 9) SIGKILL	10) SIGUSR1	11) SIGSEGV	12) SIGUSR2
13) SIGPIPE	14) SIGALRM	15) SIGTERM	16) SIGSTKFLT
... many more lines omitted ...
darkstar:~$ kill 1234 # SIGTERM
darkstar:~$ kill -s 9 1234 # SIGKILL
darkstar:~$ kill -s 1 1234 # SIGHUP
darkstar:~$ kill -s HUP 1234 # SIGHUP

Sometimes you may wish to terminate all running processes with a certain name. You can kill processes by name with killall(1). Just pass the same arguments to killall that you would pass to kill.

darkstar:~$ killall bash # SIGTERM
darkstar:~$ killall -s 9 bash # SIGKILL
darkstar:~$ killall -s 1 bash # SIGHUP
darkstar:~$ killall -s HUP bash # SIGHUP

top

cron

Sources


In Other Languages
Translations of this page?:
QR Code
QR Code slackbook:process_control (generated for current page)