ps, pstree, pgrep, kill/signals, nice/renice, job control, cgroups, ulimit, top/htop/btop.

ps

• All processes (BSD style): ps aux

• All processes (POSIX style): ps -ef

• Tree view: ps auxf / ps -ejH

• By user: ps -u jo

• By name (exact): ps -C nginx

• By PID with custom columns: ps -p 1234 -o pid,ppid,user,%cpu,%mem,stat,start,time,comm

• Custom columns + sort (desc by mem): ps -eo pid,ppid,user,%cpu,%mem,comm --sort=-%mem | head -20

• Top CPU: ps aux --sort=-%cpu | head -20

• All threads: ps -eLf

• Threads for one process: ps -T -p 1234

pstree

• Full tree: pstree

• Include PIDs: pstree -p

• Show users: pstree -u

• Command-line args: pstree -a

• Tree rooted at PID: pstree -p 1234

Finding processes

• By name (PID only): pgrep nginx

• Match full command line: pgrep -f "python script.py"

• With command line: pgrep -a nginx

• With name: pgrep -l nginx

• By user: pgrep -u jo

• Count matches: pgrep -c nginx

• What’s using a port: ss -tlnp sport = :80 / fuser 80/tcp

• What’s using a file: fuser /var/log/syslog / lsof /var/log/syslog

• Open files of process: lsof -p 1234

• Network sockets: lsof -i / lsof -i :80 / lsof -i tcp

Kill

• Graceful (SIGTERM): kill 1234

• Force (SIGKILL): kill -9 1234

• Reload config (SIGHUP): kill -HUP 1234

• By exact name: killall nginx

• By pattern (full command): pkill -f "python script.py"

• All processes of a user: pkill -u baduser

• Kill on port: fuser -k 80/tcp

• Kill process group (negative PID): kill -TERM -1234

Signals

• List signals: kill -l

• Send by name or number: kill -SIGTERM 1234 / kill -15 1234

• Trap in shell script: trap 'echo "Caught SIGINT"; cleanup; exit' INT TERM

nice / renice

Range -20 (highest) to 19 (lowest). Default 0. Negative needs root.

• Start lower priority: nice -n 10 tar czf backup.tar.gz /data

• Start higher (root): sudo nice -n -5 important-process

• Change running process: renice 10 -p 1234

• For all of a user: sudo renice 15 -u baduser

• For a process group: sudo renice 10 -g 5678

• Check value: ps -o pid,ni,comm -p 1234

Job control

• Background: long-command &

• List jobs: jobs / jobs -l (with PID)

• Suspend foreground: Ctrl+Z

• Resume in background: bg / bg %2

• Resume in foreground: fg / fg %1

• Wait for jobs: wait / wait %1 / wait 1234

• Detach from shell: disown %1 / disown -a

• Immune to hangup, output redirected: nohup long-command > output.log 2>&1 &

• Fully detached (new session): setsid long-command &

cgroups (systemd)

• View hierarchy: systemd-cgls

• Like top for cgroups: systemd-cgtop

• Set memory limit on service: sudo systemctl set-property nginx.service MemoryMax=512M

• Set CPU quota (100% = 1 core): sudo systemctl set-property myapp.service CPUQuota=50%

• Persistent via override (systemctl edit myapp.service):

  [Service]
  MemoryMax=1G
  MemoryHigh=800M
  CPUQuota=200%
  TasksMax=100
  IOWeight=50
  

• Transient scope with limits: sudo systemd-run --scope -p MemoryMax=256M -p CPUQuota=50% ./heavy-task

• Current cgroup of process: cat /proc/1234/cgroup

• Available controllers: cat /sys/fs/cgroup/cgroup.controllers

ulimit

• All limits: ulimit -a

• Open files: ulimit -n

• Max user processes: ulimit -u

• Virtual memory: ulimit -v

• Stack: ulimit -s

• Limits of running process: cat /proc/1234/limits

• Set soft (session): ulimit -n 65535

• Set hard (root): ulimit -Hn 65535

• Persistent (/etc/security/limits.conf):

  jo           soft  nofile  65535
  jo           hard  nofile  65535
  @developers  soft  nproc   4096
  

• Systemd service limit (in unit):

  [Service]
  LimitNOFILE=65535
  LimitNPROC=4096
  

Real-time monitors

• top — classic: top (M=sort mem, P=sort cpu, k=kill, r=renice, 1=per-CPU, c=full cmd)

• htop — improved: htop (F5 tree, F6 sort, F9 kill, F2 setup)

• btop — modern: btop

• iotop — per-process disk I/O: sudo iotop -o

• iftop — per-connection bandwidth: sudo iftop -i enp0s3

• nethogs — per-process bandwidth: sudo nethogs enp0s3