nohup, disown, and screen: Keeping Processes Running After Logout

nohup, disown, and screen: Keeping Processes Running After Logout
Reading time: About 12 min Difficulty: Intermediate

What You'll Learn

  • Why processes die when you close a terminal or SSH session
  • How to use nohup, disown, and screen to keep processes alive after logout
  • When to reach for each tool

Quick Reference

Situation Tool
Start a command that survives logout nohup cmd &
Detach an already-running job disown %1
Keep a full session and reattach later screen or tmux

Why Do Processes Die When You Log Out?

When you close a terminal or drop an SSH connection, the shell sends SIGHUP (Signal 1: Hangup) to every process in its job table. The default signal handler exits, so background jobs die along with the shell.

$ sleep 300 &
[1] 12345
$ exit   # shell exits → SIGHUP sent to sleep → PID 12345 terminates

The three tools below each cut the SIGHUP delivery path at a different layer.

nohup — Use Before Starting

nohup ("no hangup") starts a command with SIGHUP ignored from the very beginning.

$ nohup long-job.sh &
  • & is optional but idiomatic — background execution is the usual intent
  • STDOUT and STDERR are automatically redirected to nohup.out in the current directory
  • Specify a custom log file to avoid nohup.out growing unbounded:
$ nohup ./deploy.sh > /var/log/deploy.log 2>&1 &

Discard output entirely when you don't need it:

$ nohup ./job.sh > /dev/null 2>&1 &

nohup only blocks SIGHUP — it does not fully daemonize the process. After the shell exits, the process stays alive but gets reparented to init (PID 1).

disown — Use After Starting

disown removes a job from the shell's job table so SIGHUP is no longer delivered when the shell exits. Use it when you forgot to prefix with nohup.

# Check running jobs
$ jobs
[1]+  Running    long-job.sh &

# Remove job 1 from the job table
$ disown %1

# Remove all jobs
$ disown -a

disown -h marks the job to not receive SIGHUP without removing it from the job list — you can still see it with jobs.

$ disown -h %1

Recovering When You Forgot nohup

# 1. If the job is in the foreground, suspend it first
Ctrl+Z
[1]+  Stopped    ./deploy.sh

# 2. Resume it in the background
$ bg %1
[1]+ ./deploy.sh &

# 3. Cut SIGHUP delivery
$ disown %1

# 4. Safe to close the terminal

screen — Persist the Entire Session

screen is a terminal multiplexer that virtualizes the entire terminal session. The killer feature is detach and reattach — disconnect, close your laptop, reconnect from a different machine, and find your session exactly where you left it.

# Create a named session and enter it
$ screen -S mysession

# Do your work inside screen
# Detach without killing (screen stays alive in background)
Ctrl+A D

# Reattach from another terminal or SSH session
$ screen -r mysession

# List all sessions
$ screen -ls

Common screen Keybindings

Action Key
Detach Ctrl+A D
New window Ctrl+A C
Next window Ctrl+A N
Previous window Ctrl+A P
End session exit in the last window

screen vs tmux

tmux offers more features and a more flexible config system. If tmux is available on your server, prefer it. See Getting Started with tmux for a comparison. Use screen when tmux is unavailable or you want the simplest possible setup.

Choosing the Right Tool

Tool When Notes
nohup cmd & Before starting Simple, built-in, output goes to nohup.out
disown After starting Forgot nohup — no log file created automatically
screen / tmux Either Detach/reattach, multiple windows, session sharing

Common mistakes to avoid

  • Running a background job without nohup and then closing the terminal — SIGHUP kills it
  • Pressing Ctrl+D inside a screen window thinking you "closed" screen — you ended the session; use Ctrl+A D to detach instead
  • Letting nohup.out grow without bound — always specify a log path or redirect to /dev/null

Next Reading