Solution
It was found (here, and here) that Podman uses its own DNS server, aardvark-dns
which is bound to port 53 (this explains why I was able to bind to 53 with nc
on the host while the container would still fail). So the solution is to bridge the network for that port. So, in the compose file, the ports section would become:
ports:
- "<host-ip>:53:53/tcp"
- "<host-ip>:53:53/udp"
- "80:80/tcp"
where <host-ip>
is the ip of the machine running the container — e.g. 192.168.1.141
.
Original Post
I so desperately want to bash my head into a hard surface. I cannot figure out what is causing this issue. The full error is as follows:
Error: cannot listen on the UDP port: listen udp4 :53: bind: address already in use
This is my compose file:
version: "3"
services:
pihole:
container_name: pihole
image: docker.io/pihole/pihole:latest
ports:
- "53:53/tcp"
- "53:53/udp"
- "80:80/tcp"
environment:
TZ: '<redacted>'
volumes:
- './etc-pihole:/etc/pihole'
- './etc-dnsmasq.d:/etc/dnsmasq.d'
restart: unless-stopped
and the result of # ss -tulpn
:
Netid State Recv-Q Send-Q Local Address:Port Peer Address:Port Process
udp UNCONN 0 0 [fe80::e877:8420:5869:dbd9]:546 *:* users:(("NetworkManager",pid=377,fd=28))
tcp LISTEN 0 128 0.0.0.0:22 0.0.0.0:* users:(("sshd",pid=429,fd=3))
tcp LISTEN 0 128 [::]:22 [::]:* users:(("sshd",pid=429,fd=4))
I have looked for possible culprit services like systemd-resolved
. I have tried disabling Avahi. I have looked for other potential DNS services. I have rebooted the device. I am running the container as sudo (so it has access to all ports). I am quite at a loss.
- Raspberry Pi Model 1 B Rev 2
- Raspbian (bookworm)
- Kernel v6.6.20+rpt-rpi-v6
- Podman v4.3.1
- Podman Compose v1.0.3
EDIT (2024-03-14T22:13Z)
For the sake of clarity, # netstat -pna | grep 53
shows nothing on 53, and # lsof -i -P -n | grep LISTEN
shows nothing listening to port 53 — the only listening service is SSH on 22, as expected.
Also, as suggested here, I tried manually binding to port 53, and I was able to without issue.
Can you use nc to bind to udp 53 yourself?
Yup. I ran
# nc -u -l 0.0.0.0 53
to listen on port 53. Then I ran# drill @127.0.0.1 53 archlinux.org
in another shell. I saw the request in the listening shell.Hmm. Can you run another service like https://github.com/vhiribarren/docker-echo-server on udp 53? If it fails, can you bind another port, both under and over 1024, with success? That would tell us if it’s something with that image or a system problem.
Edit: I remember when I was running pihole under docker, maybe this? Not sure if anyone has mentioned it: https://github.com/pi-hole/docker-pi-hole/issues/968 (the NET_ADMIN capability is different from running as root).
Maybe also try starting the container without the port maps in the compose file to see if that works?
Also what network are you using? I think I had to put mine in host networking.
See the post for the solution.