Processes

Jobs

Use job control to run and manage multiple processes from a single shell session:

CTRL + z            # suspend the current process and send it to the background
jobs                # list all background jobs
fg                  # bring the most recent job to the foreground
fg 1                # bring job 1 to the foreground
bg 1                # resume job 1 in the background
<program> &         # start <program> directly in the background

ps

Use ps to list processes running on your system.

The STAT column uses single-letter codes to show a process’s current state. S (interruptible sleep) means the process is waiting for an event and can be interrupted by signals. D (uninterruptible sleep) means the process is waiting (usually on I/O) and cannot be interrupted. Z (defunct) marks a zombie process waiting for its parent to clean it up.

TTY stands for teletypewriter, historically a physical terminal connected to a mainframe. Today it’s the terminal connected to your machine locally or over SSH. pts/0 is a pseudo-terminal (a virtual terminal assigned by a terminal emulator):

ps                                  # list processes for the current user
pidof <name>                        # get the PID of a named process
ps a                                # list processes for all users
ps au                               # list processes with usernames
ps aux                              # list all processes, including daemons not tied to a TTY
ps -ef                              # every process, from init down
ps aux | grep <keyword>             # search processes by keyword
ps u -C <keyword>                   # show processes matching keyword
ps aux --sort=-pcpu                 # sort by CPU usage (highest first)
ps aux --sort=-pcpu | head -5       # show the 5 processes using the most CPU
ps aux --sort=-pmem | head -5       # show the 5 processes using the most memory

pstree

Use pstree to visualize running processes as a tree, showing parent/child relationships at a glance. The -p flag includes each process’s PID:

pstree -p
systemd(1)─┬─ModemManager(789)─┬─{ModemManager}(805)
           │                   ├─{ModemManager}(806)
           │                   └─{ModemManager}(813)
           ├─cron(857)
           ├─dbus-daemon(717)
           ├─login(866)───bash(982)
           ├─multipathd(364)─┬─{multipathd}(376)
           │                 ├─{multipathd}(377)
           │                 ├─{multipathd}(378)
           │                 ├─{multipathd}(379)
           │                 ├─{multipathd}(380)
           │                 └─{multipathd}(381)
...

Change process priority

Linux assigns each process a priority that determines how much CPU time the scheduler gives it. Process priority changes are most common in CPU-intensive workloads like data processing or machine learning jobs.

nice values range from -20 (highest priority) to 19 (lowest priority). The higher the nice value, the more a process yields to others. nice sets the priority of a process when you start it. renice changes the priority of a process that is already running. You cannot change the niceness of a process you do not own, and you need sudo to lower the nice value (increase priority) of any process.

nice

Pass -n with a value from -20 to 19 to set the priority when starting a process:

nice -n <num> <program>

renice

renice changes the priority of a process that is already running. The -n flag takes an absolute niceness (NI) value, replacing the current value rather than adjusting it. New processes start with a default NI of 0:

renice -n <num> -p <PID>
renice -n <num> <process-name>

In this example, vim is running with PID 12047 at the default NI of 0. Raising its NI to 10 reduces its CPU scheduling priority.

1. Check the current PRI and NI values with ps -l.

   ps -l
   F S   UID     PID    PPID  C PRI  NI ADDR SZ WCHAN  TTY          TIME CMD
   0 S  1000    1093    1091  0  80   0 -  2254 do_wai pts/0    00:00:03 bash
   0 T  1000   12047    1093  0  80   0 -  8981 do_sig pts/0    00:00:00 vim
   0 R  1000   12196    1093 99  80   0 -  2729 -      pts/0    00:00:00 ps
   

2. Set the NI value for the vim process (PID 12047) to 10.

   renice -n 10 -p 12047
   12047 (process ID) old priority 0, new priority 10
   

3. Confirm the updated PRI and NI values.

   ps -l
   F S   UID     PID    PPID  C PRI  NI ADDR SZ WCHAN  TTY          TIME CMD
   0 S  1000    1093    1091  0  80   0 -  2254 do_wai pts/0    00:00:03 bash
   0 T  1000   12047    1093  0  90  10 -  8981 do_sig pts/0    00:00:00 vim
   0 R  1000   12205    1093 99  80   0 -  2729 -      pts/0    00:00:00 ps
   

Misbehaving processes

The kill command sends signals to processes. A signal is a notification from the kernel, another process, or a user that tells a process to take action. SIGTERM (15) asks a process to stop gracefully. SIGKILL (9) forces it to stop immediately, bypassing any cleanup handlers. Reserve it for processes that do not respond to SIGTERM, such as zombie processes. For a full signal reference, run man 7 signal:

kill <PID>              # send SIGTERM
killall <process>       # send SIGTERM to all processes matching <process>

kill -9 <PID>           # send SIGKILL
killall -9 <process>    # not recommended

Managing system processes

Daemons run in the background and typically start at boot. When you install an app to run as a service, the system configures it to start automatically. The terms service, daemon, and unit are interchangeable.

systemd is the init system (PID 1) responsible for managing these services. Manage units with systemctl. To check whether a unit starts automatically after installation, look for the vendor preset field in the Loaded line of systemctl status <unit>.

systemctl

The following commands cover the most common unit management tasks:

systemctl                       # list all available units
systemctl list-units            # same as above
systemctl status <unit>         # show status, enabled state, and recent logs

systemctl start <unit>          # start the unit
systemctl stop <unit>           # stop the unit
systemctl restart <unit>        # stop then start the unit
systemctl reload <unit>         # reload configuration without stopping the unit (not supported by all units)

systemctl enable <unit>         # start the unit automatically at boot
systemctl disable <unit>        # do not start the unit at boot
systemctl enable --now ssh      # enable a unit and start it immediately

init

init is the first process the kernel starts at boot, always assigned PID 1. On modern Linux systems, /sbin/init is a symlink to systemd. These commands let you confirm that relationship:

ps -ef | grep init              # find the init process
file /sbin/init                 # show what init points to

Scheduling with cron

cron runs commands on a schedule. Each user has a personal schedule file called a crontab (cron table), where each line defines one job. Cron jobs require fully-qualified command paths. Find the path for a command with which. root can also schedule system-wide tasks.

Each crontab entry follows this format:

<minute> <hour> <day-of-month> <month> <day-of-week> <command>

An asterisk (*) in any field means “every.” Values range from 0–59 for minutes, 0–23 for hours, 1–31 for day of month, 1–12 for month, and 0–7 for day of week (0 and 7 both represent Sunday):

crontab -l                              # view your crontab
sudo crontab -u <user> -l               # view another user's crontab

crontab -e                              # open your crontab in the default editor
EDITOR=vim crontab -e                   # open your crontab in vim for this invocation

# --- Example crontab entries --- #
3 0 * * 4 /usr/local/bin/backup.sh      # run backup.sh at 12:03 AM each Thursday
0 0 * * * /usr/bin/apt update           # run apt update once at midnight each night