"Address already in use": find who is using port 80 (or any port)

You start nginx and it refuses:

$ sudo systemctl start nginx
Job for nginx.service failed because the control process exited with error code.
See "systemctl status nginx.service" and "journalctl -xeu nginx.service" for details.

The log has the real message:

nginx: [emerg] bind() to 0.0.0.0:80 failed (98: Address already in use)
nginx: [emerg] still could not bind()

Only one program can listen on a given address and port. Something else already holds port 80, so the operating system refuses to give it to nginx. The fix is always the same: find who holds the port, then decide which of the two should have it.

The same problem has a different message in every program:

Error: listen EADDRINUSE: address already in use :::3000            # Node.js
OSError: [Errno 98] Address already in use                          # Python
listen tcp :8080: bind: address already in use                      # Go
Bind for 0.0.0.0:80 failed: port is already allocated               # Docker

TL;DR

sudo ss -tlnp | grep ':80 '                 # who listens on port 80?
sudo lsof -i :80 -sTCP:LISTEN               # the same, with lsof
sudo systemctl disable --now apache2        # stop it now and at boot (httpd on RHEL)
sudo systemctl start nginx

1. Read the error

sudo journalctl -u nginx -n 20 --no-pager

Note the address and port from the bind() line — 0.0.0.0:80 is port 80 on all IPv4 addresses, [::]:80 is port 80 on all IPv6 addresses. Error 98 means the port is taken. If you see error 13 (Permission denied) instead, it’s a different problem — see the end of this article.

2. Find the process that holds the port

sudo ss -tlnp | grep ':80 '

LISTEN 0      511    0.0.0.0:80         0.0.0.0:*         users:(("apache2",pid=812,fd=6))

The options: -t TCP, -l listening sockets only, -n numbers instead of names (80, not http), -p the process. The last column is the answer: apache2, process ID 812.

Use sudo. Without it, ss still shows that the port is taken, but not who has it if the process belongs to another user — the users:(…) column is simply empty.

lsof gives the same answer, if it’s installed:

sudo lsof -i :80 -sTCP:LISTEN

COMMAND PID USER FD TYPE DEVICE SIZE/OFF NODE NAME
apache2 812 root 4u IPv4 21534  0t0      TCP  *:http (LISTEN)

The old netstat -tlnp also works where it’s still installed; on current distributions it’s usually missing, and ss replaces it.

3. Find out where it comes from

A process name and PID are not enough to fix it for good — you need to know what starts it. Ask systemd which unit the PID belongs to:

systemctl status 812

The first line names the unit, for example apache2.service. If the process is not part of any service (you see something like session-3.scope), someone started it by hand from a shell — ps -fp 812 shows the full command and its parent.

The usual suspects:

4. Fix it

Another service has the port and you don’t need it. Stop it and keep it from starting at boot — otherwise the problem comes back after the next reboot, when the two services race for the port:

# Ubuntu / Debian
sudo systemctl disable --now apache2

# RHEL / Rocky / AlmaLinux
sudo systemctl disable --now httpd

sudo systemctl start nginx

If you are sure nothing needs the other web server, remove the package (sudo apt remove apache2 / sudo dnf remove httpd).

You need both services. Move one of them to a different port — for example Apache to 8080 (Listen 8080 in /etc/apache2/ports.conf on Debian/Ubuntu, /etc/httpd/conf/httpd.conf on RHEL) — and put nginx in front of it as a reverse proxy.

nginx was started by hand. systemd’s nginx can’t start because “your” nginx already has the port. Stop the stray one, then start it the normal way:

sudo nginx -s stop
sudo systemctl start nginx

A container has the port. Find it and stop it, or publish it on another port:

docker ps --filter publish=80

A process with no service. End it with kill:

sudo kill 812        # asks the process to exit (SIGTERM)
sudo kill -9 812     # forces it, only if it ignores the first one

Then check again with ss that the port is free before you start your service.

5. When nginx collides with itself

Sometimes ss shows the port is free, yet nginx still fails with error 98. Then nginx’s own configuration asks for the same address twice. A typical case: one server block has listen [::]:80 ipv6only=off; (IPv6 and IPv4), and another has listen 80; (IPv4). The second bind() collides with the first.

Find every listen line nginx uses:

sudo nginx -T 2>/dev/null | grep -n 'listen'

Keep IPv4 and IPv6 separate (listen 80; and listen [::]:80;, without ipv6only=off) and check the config before you start again:

sudo nginx -t && sudo systemctl start nginx