Port Already in Use Linux: How to Find and Fix It

bind: address already in use. Service won't start. Something is holding the port. Here's every way to find it and fix it.

TL;DR

# Find what's using port 8080
ss -tlnp | grep :8080

# Kill it
fuser -k -TERM 8080/tcp

# Or explicitly
kill -15 <pid>

Find What's Using the Port

Method 1: ss (fastest)

ss -tlnp | grep :8080
# LISTEN 0 128 0.0.0.0:8080 0.0.0.0:* users:(("node",pid=4521,fd=12))

The pid=4521 is what you need.

Method 2: lsof

lsof -i :8080
# COMMAND  PID     USER   FD   TYPE NODE NAME
# node    4521  appuser  12u  IPv4  TCP *:http-alt (LISTEN)

Method 3: fuser

fuser 8080/tcp
# 4521   ← PID only

Kill the Process

# Graceful (always try this first)
kill -15 4521

# Confirm it's gone
sleep 3 && ss -tlnp | grep :8080

# Force kill only if graceful didn't work
kill -9 4521

One-liner with fuser:

fuser -k -TERM 8080/tcp   # SIGTERM
fuser -k 8080/tcp         # SIGKILL (if still alive)

If It's a systemd Service

Don't kill the PID — stop the service:

# Find which service owns the process
systemctl status $(ss -tlnp | grep ':8080' | grep -oP 'pid=\K[0-9]+' | head -1)

# Stop the service properly
systemctl stop myapp

# Now start your service
systemctl start newapp

Killing a PID when systemd manages it will cause an immediate restart.

The TIME-WAIT Case

Sometimes nothing is listening on the port but you still get "address already in use":

ss -tlnp | grep :8080
# Empty — nothing listening

ss -tn state time-wait | grep :8080
# TIME-WAIT socket still exists

TIME-WAIT sockets linger for ~60 seconds after a connection closes. Most servers set SO_REUSEADDR so this shouldn't block binding — but if your app doesn't, you have two options:

# Option 1: wait 60 seconds and retry

# Option 2: enable SO_REUSEADDR in your app (app-level fix)

# Option 3: shrink TIME-WAIT duration (sysctl — use carefully)
sysctl net.ipv4.tcp_fin_timeout
# Default: 60
sudo sysctl -w net.ipv4.tcp_fin_timeout=30

Real Examples

New app won't start after deploy

ss -tlnp | grep :3000
# node  9901  ...  (old process still running)

# Kill old process
kill -15 9901
sleep 2

# Start new version
systemctl start myapp

Two services configured to the same port

ss -tlnp | grep :80
# nginx: 1234
# apache2: 5678  ← both trying port 80

# Fix: change one to a different port, or remove one
systemctl stop apache2
systemctl disable apache2

Docker container port conflict

docker run -p 8080:8080 myimage
# Error: Bind for 0.0.0.0:8080 failed: port is already allocated

# Find what Docker container has the port
docker ps | grep 8080
# or
ss -tlnp | grep :8080   # shows dockerd or the container process

# Stop the conflicting container
docker stop <container_name>
docker run -p 8080:8080 myimage

Prevent It Recurring

Use systemd socket activation

Instead of the app binding to the port directly, let systemd manage the socket:

# /etc/systemd/system/myapp.socket
[Socket]
ListenStream=8080
BindIPv6Only=both

[Install]
WantedBy=sockets.target

This way the port is held by systemd, and a restart of the app doesn't release and re-acquire it — eliminating the race condition window.

Check on startup

# Add to your start script
if ss -tlnp | grep -q ':8080'; then
  echo "ERROR: port 8080 already in use"
  ss -tlnp | grep :8080
  exit 1
fi

Common Mistakes

Mistake 1: kill -9 immediately SIGTERM first. SIGKILL doesn't let the process clean up — can leave temp files, locks, or corrupt state.

Mistake 2: Killing a PID managed by systemd It respawns immediately. Use systemctl stop.

Mistake 3: Confusing TIME-WAIT for a live listener ss -tlnp shows listeners only. If it's empty, check ss -tn state time-wait. Two completely different problems.

Conclusion

ss -tlnp | grep :<port> → find the PID → kill -15 <pid> or systemctl stop <service>. If nothing shows but bind still fails, check for TIME-WAIT sockets. If systemd manages the service, always use systemctl stop, not kill.

Related: Linux Kill Process by Port — all methods to kill by port. Check Open Ports in Linux: ss vs netstat — full ss usage guide.