Compose domains not generating Traefik config → all routes return 404

[xndyz]

To Reproduce

To Reproduce

1. Deploy a Docker Compose app with two services:
   • client running on port 3002
   • backend running on port 3001
2. Configure domains in Dokploy:
   • example.com → service client → path / → port 3002
   • example.com → service backend → path /api → port 3001
3. Redeploy the compose
4. Check Traefik dynamic config directory:

ls /etc/dokploy/traefik/dynamic
5. Access:
`https://example.com

https://example.com/api/config`

Current vs. Expected behavior

Expected:

• Dokploy generates a dynamic Traefik config file for the compose
• Routes:
  • / → client (3002)
  • /api → backend (3001)
• App loads correctly

Actual:

• No dynamic config file is created for this compose
• /etc/dokploy/traefik/dynamic does not contain any file for this app
• All routes return 404
• Containers are running and healthy

Provide environment information

Ubuntu 20.04

Which area(s) are affected? (Select all that apply)

Docker Compose

Are you deploying the applications where Dokploy is installed or on a remote server?

Same server where Dokploy is installed

Additional context

No response

Will you send a PR to fix it?

No

[dosubot]

Hi @xndyz! I'm Dosu and I’m helping the dokploy team.

@xndyz This is expected behavior — Docker Compose services in Dokploy use label-based routing, not file-based config. Unlike standalone applications, compose deployments don't generate YAML files in /etc/dokploy/traefik/dynamic/. Instead, Traefik labels are injected directly into your docker-compose.yml and read by Traefik's Docker provider from running containers [1].

So the absence of config files in that directory is by design and not the cause of your 404s.

The 404s are most likely caused by one of these:

1. Missing dokploy-network connectivity — Compose containers must be connected to dokploy-network for Traefik to discover them. This was a critical bug fixed in PR refactor(network-service): enhance network addition logic to include … #3614 [2] [3]. If you're on an older version, your containers may be isolated on their project network and invisible to Traefik. Make sure you're running the latest Dokploy version.

2. Redeploy required after domain changes — Since labels live on running containers, any domain configuration change requires a full redeploy (container recreation) to take effect [4]. Simply saving domain settings in the UI isn't enough.

3. Healthcheck failures — If your compose services have a healthcheck definition and it's failing, Traefik will silently refuse to create routers for those services, even if labels are correct [5]. You can verify with docker ps — check that containers show as (healthy), not (unhealthy) or (health: starting).

To debug further:

• Run docker inspect <container_name> | grep -A 50 Labels on your client/backend containers to verify Traefik labels are present
• Check that your containers are connected to dokploy-network: docker network inspect dokploy-network
• Review Traefik logs for any label parsing errors: docker logs dokploy-traefik

What version of Dokploy are you running?

To reply, just mention @dosu.

---

Docs are dead. Just use Dosu.

[siskomx]

We hit this exact symptom and root-caused it. Sharing in case it helps the maintainers and others tracking this.

Trigger: the host's Docker Engine got auto-upgraded from v27 → v29.4.1 (Ubuntu unattended-upgrades). After the upgrade, every Compose-type Dokploy app started returning 404, while Application-type apps kept working. Containers themselves were healthy and reachable from inside dokploy-network.

Root cause: Traefik v3.x (Dokploy v0.29.2 ships v3.1.2; we also tested v3.5) hardcodes client.WithVersion("1.24") in its embedded Docker SDK as a "compatibility default" for old Docker daemons. Docker 28.0 raised MinAPIVersion to 1.44, and Docker 29 keeps it at 1.40. So the daemon now rejects every API call from Traefik:

ERR Failed to retrieve information of the docker client and server host
error="Error response from daemon: client version 1.24 is too old.
Minimum supported API version is 1.40, please upgrade your client to a newer version"
providerName=docker
Both the docker and swarm providers fail identically — they share the same client. Traefik's file provider keeps working, which is why Application-type apps (which have generated yml files in /etc/dokploy/traefik/dynamic/) appeared fine, while Compose-type apps (which rely entirely on container labels via the docker provider) returned 404 with no obvious clue.

Things we tried that did not help:

Upgrading Traefik to v3.5 — same v1.24 hardcode.
Setting DOCKER_API_VERSION=1.44 env var on Traefik — Traefik v3.x builds its docker client without client.FromEnv, so the env var is ignored.
tecnativa/docker-socket-proxy between Traefik and dockerd — the proxy passes the /v1.24/... URL through unchanged; dockerd still rejects.
Workaround that worked: downgrade Docker Engine to v27.x and pin it.

apt-get install -y --allow-downgrades
docker-ce=5:27.5.1-1ubuntu.24.04noble
docker-ce-cli=5:27.5.1-1ubuntu.24.04noble
docker-ce-rootless-extras=5:27.5.1-1ubuntu.24.04noble
apt-mark hold docker-ce docker-ce-cli docker-ce-rootless-extras containerd.io
Docker 27 still accepts API v1.24. After the downgrade, Compose apps' container labels load via the docker provider as expected and routing works again.

Suggested fixes for Dokploy upstream:

(Short-term) Document the Docker 27 pin in the install docs, or have the installer set apt-mark hold automatically — many users will hit this as their hosts auto-upgrade.
(Better) Generate dynamic yml files for Compose apps too, mirroring how Application-type apps work. That sidesteps the docker provider entirely and makes Compose routing resilient to Docker SDK bugs in Traefik. We tested this manually (writing a yml file mirroring the existing container labels) and it works perfectly.
(Best, if Traefik upstream cooperates) Track traefik/traefik#XXXX — Traefik should call cli.NegotiateAPIVersion() on its docker client, or expose an apiVersion config option, so Docker 28+ stops being a hard wall.
Environment for reference: Docker 29.4.1 (API 1.54, MinAPIVersion 1.40), Ubuntu 24.04 noble, Dokploy v0.29.2, Traefik v3.1.2 (also tested v3.5).

[jsheely]

It looks like this may be fixed already in Traefik 3.6 if I followed this thread correctly traefik/traefik#12256

So may just be able to bump the version of traefik that dokploy is using to resolve the problem as well?

[jsheely]

The dokploy install script also references Traefik 3.6.7

 292   │     docker run -d \
 293   │         --name dokploy-traefik \
 294   │         --restart always \
 295   │         -v /etc/dokploy/traefik/traefik.yml:/etc/traefik/traefik.yml \
 296   │         -v /etc/dokploy/traefik/dynamic:/etc/dokploy/traefik/dynamic \
 297   │         -v /var/run/docker.sock:/var/run/docker.sock:ro \
 298   │         -p 80:80/tcp \
 299   │         -p 443:443/tcp \
 300   │         -p 443:443/udp \
 301   │         traefik:v3.6.7

[jsheely]

So a little more rolling up the sleeves and I determined that things in the install script have changed from when I originally installed dokploy and the two are not compatible.

Specifically how the database is created with a docker secrets password.

After a little bit of a scare of re-installing everything to find out I had no projects. I was able to find the old location volume name that dokploy postgres was created under.

I'll try and itemize my steps better later but basically it involved these things

• Downloading the install script and hard coding the password it generates so its consistent to my old password curl -sSL https://dokploy.com/install.sh -o install.sh
• Copying the database files from the old location to the new path sudo cp -a /mnt/data/docker/var-lib-docker/volumes/dokploy-postgres-database/_data/. /mnt/data/docker/var-lib-docker/volumes/dokploy-postgres/_data
• Removing all the services and traefik container.
• Reinstalling with the modified install script ./install.sh

Everything was back up and running with the latest traefik container and upgraded postgres, dokploy.

[andreaswiren]

Good findings here, we struck this as well. No dynamic files generated and the labels are not found either.

Traefik log shows,

2026-05-11T21:21:44Z ERR Failed to list containers for docker error="Get \"http://%2Fvar%2Frun%2Fdocker.sock/v1.51/containers/json?all=1\": context canceled" providerName=docker 2026-05-11T21:46:58Z ERR Failed to list services for docker swarm mode error="Get \"http://%2Fvar%2Frun%2Fdocker.sock/v1.51/services\": context canceled" providerName=swarm 2026-05-11T21:46:58Z ERR Failed to list containers for docker error="Get \"http://%2Fvar%2Frun%2Fdocker.sock/v1.51/containers/json?all=1\": context canceled" providerName=docker

Manually creating dynamic-files for now as this is a small install. I would also prefer having dynamic yml created for compose as well to get dockerfiles and compose's to behave more the same.

OS
Ubuntu 24.04.4 LTS

[anhthinhvup]

Because the containers are healthy but Traefik never gets the dynamic config, this looks like a routing-generation
failure rather than an app bug. If you want, I can help narrow down whether the compose metadata, domain mapping, or
generator path is breaking here. If you want to continue privately, email me at phamvanthinhcontact2004@gmail.com and
I will send the intake checklist.

[JuanJoCasamitjana]

I can confirm, downgrading Docker from 28+ to 27.x solves the problem. I've been experiencing this problem for a couple of weeks after an OS update. So definitely is not the dynamic generation of traefik files but rather the incompatibility between Traefik and Docker.

OS: Fedora Linux 42 (Adams) x86_64
Kernel: Linux 6.19.14-100.fc42.x86_64

[ISCIrvingDev]

I'm experiencing the exact same issue on Dokploy v0.29.4. After extensive debugging, I can confirm the root cause and provide a workaround.

Diagnosis

When configuring a domain for a Docker Compose service in the Dokploy UI:

1. Dokploy correctly injects Traefik labels into the container (via its overlay on the compose file). Verified with docker inspect:

   • traefik.enable=true
   • traefik.http.routers.<name>.rule=Host(\domain`)`
   • traefik.http.services.<name>.loadbalancer.server.port=<port>
   • etc.

2. Dokploy correctly connects the container to dokploy-network. Verified with docker inspect showing the container on both api-network and dokploy-network.

3. But Dokploy does NOT generate the Traefik dynamic config file in /etc/dokploy/traefik/dynamic/. This is the critical failure:

$ grep -r "my-app" /etc/dokploy/traefik/dynamic/
# (no results)

4. Traefik uses the file provider (not the Docker provider) for routing in Dokploy. So while the labels exist on the container, they are effectively ignored. Only services with a corresponding .yml file in /etc/dokploy/traefik/dynamic/ get routed correctly.

Evidence

All working services in my Dokploy instance have their corresponding files:

$ ls /etc/dokploy/traefik/dynamic/
app-1-xxxxxx.yml  app2-xxxxxx.yml  app-3-xxxxxx.yml  ...

Each of these files was generated by Dokploy for application-type deployments (not compose). The compose deployment is the only one missing its Traefik config file.

Workaround

Manually create the Traefik dynamic config file:

cat > /etc/dokploy/traefik/dynamic/<service-name>.yml << 'EOF'
http:
  routers:
    <router-name>-web:
      rule: Host(`<domain>`)
      service: <service-name>-web
      middlewares:
        - redirect-to-https
      entryPoints:
        - web
    <router-name>-websecure:
      rule: Host(`<domain>`)
      service: <service-name>-websecure
      middlewares: []
      entryPoints:
        - websecure
      tls:
        certResolver: letsencrypt
  services:
    <service-name>-web:
      loadBalancer:
        servers:
          - url: http://<container-hostname>:<port>
        passHostHeader: true
    <service-name>-websecure:
      loadBalancer:
        servers:
          - url: http://<container-hostname>:<port>
        passHostHeader: true
EOF

Traefik picks up the file automatically (hot-reload, no restart needed).

Environment

• Dokploy version: v0.29.4
• OS: Ubuntu
• Deploy target: Same server where Dokploy is installed
• Affected area: Docker Compose

[JuanJoCasamitjana]

After further testing I found and confirm that the permanent solution is to just upgrade the Traefik container to v3.6.7:
Docker: 29.5.2, build 79eb04c
OS: Fedora Linux 42 (Adams) x86_64
Kernel: 6.19.14-100.fc42.x86_64
Dokploy: v0.29.4
traefik:v3.6.7

Deployed docker compose with correct domain configured, no dynamic file is generated, service is correctly routed solving the 404 error