Dokploy Dashboard Not Loading on Port 3000: The Six Causes and Their Fixes

A Dokploy dashboard that will not load is almost always one of six things, and five of them are fixed from an SSH session in a few minutes. This guide follows the order in which they occur in practice, firewall first and a rebuilt service last, using the commands in Dokploy's own instance troubleshooting page (checked 31 October 2026). Your deployed applications are probably still running while you work: the dashboard is a control plane, and Traefik keeps serving your sites without it.

1. The port is closed or you are on the wrong address

Fresh installs listen on port 3000 on every interface. If UFW is enabled and you never allowed 3000, or you already moved the dashboard behind Traefik on a domain and then closed 3000 as the hardening guide recommends, the raw IP will refuse. Check from the server:

sudo ufw status
ss -ltnp | grep 3000

If the port is listening but UFW blocks it, either open it temporarily (sudo ufw allow 3000/tcp) or use the domain you configured. If you are behind a cloud firewall or a corporate network, test from another network first. Dokploy's hardening guide says the UI should be reached over a domain through Traefik or over a VPN, so once the dashboard works again, close 3000 rather than leaving it open.

2. The disk is full

The first cause the Dokploy docs list: a full disk puts the Dokploy PostgreSQL database into recovery mode and the UI stops responding. Builds and images are the usual culprit.

df -h /
docker system df

If the root filesystem is at or near 100 percent:

docker system prune -a
docker builder prune -a
docker image prune -a

These remove stopped containers, unused images and build cache; they do not touch named volumes, so databases and uploads are safe. Then restart the Dokploy service (step 3) if the UI does not recover on its own. Afterwards turn on Dokploy's scheduled Docker cleanup and size the disk for growth; our requirements guide explains why 30 GB fills quickly.

3. The core containers did not start in order

After a reboot, Dokploy can come up before its database and log ENOTFOUND dokploy-postgres. List the core services first:

docker ps

You should see dokploy, dokploy-postgres and dokploy-traefik (installs older than v0.29.9 also run dokploy-redis; Redis is no longer used in self-hosted Dokploy since that version). Read the logs of whichever is missing or restarting:

docker service logs dokploy
docker service logs dokploy-postgres
docker logs dokploy-traefik

For the database race, scale Dokploy down and back up so it reconnects:

docker service scale dokploy=0
docker service scale dokploy=1

Give it thirty seconds and reload the dashboard.

4. Traefik has a broken dynamic configuration

If docker logs dokploy-traefik shows field not found, node: passHostHeader, a dynamic config file under /etc/dokploy/traefik has passHostHeader in the wrong place. Restart Traefik first:

docker restart dokploy-traefik

If the error persists, edit the file so that passHostHeader sits under loadBalancer, not inside the servers list:

services:
  dokploy-service-app:
    loadBalancer:
      servers:
        - url: http://dokploy:3000
      passHostHeader: true

Then restart Traefik again. This affects the dashboard only when you reach it through a domain; on the raw port 3000 Traefik is not in the path.

5. Services cannot reach each other (IPVS missing)

On minimal or custom kernels (ZimaOS, Buildroot images, some containers-as-hosts), Swarm's default VIP networking needs IPVS modules that are absent. Symptoms: the UI does not load on 3000, containers resolve names but connections hang. Check:

sudo modprobe ip_vs && lsmod | grep ip_vs

If that fails, switch the core services to DNS round-robin mode:

docker service update --endpoint-mode dnsrr dokploy-postgres
docker service update --endpoint-mode dnsrr dokploy

For a fresh install the equivalent is curl -sSL https://dokploy.com/install.sh | sudo ENDPOINT_MODE=dnsrr sh. This does not apply to a standard Ubuntu or Debian image such as the ones MassiveGRID provisions, where IPVS is present.

6. Recreate the core services

If a service is gone or will not start, the Dokploy docs provide the exact docker service create commands to recreate dokploy-postgres, dokploy-traefik and dokploy. Two rules before you run them: back up first (recreating services does not delete volumes, but a backup costs nothing), and on installations older than v0.29.3 create the auth secret the new service expects:

openssl rand -hex 32 | docker secret create dokploy_auth_secret -

Then remove and recreate the service with the command from the docs page, which binds the Docker socket, /etc/dokploy and the dokploy volume, publishes port 3000 in host mode and pins the service to a manager node. On Proxmox LXC containers add --endpoint-mode dnsrr to each create command. If you are not comfortable with that step, this is exactly the kind of server-level issue MassiveGRID's 24/7 support helps with on the Dokploy hosting plans.

Prevention

• Disk: 64 GB or more, Dokploy's scheduled cleanup on, an alert at 80 percent.
• Backups: instance backups under Web Server, Backups to an S3 destination, so a broken dashboard can be restored to a fresh server in minutes.
• Access: the dashboard on a domain behind Traefik with 2FA, port 3000 closed, so a "not loading" report is never a firewall surprise.
• Hosting: a server that migrates to healthy hardware when the host fails, so the sixth cause stays rare.

Frequently Asked Questions

Are my applications down while the Dokploy dashboard is unreachable?

Usually not. Applications run as Docker Swarm services behind Traefik, which keeps routing and renewing certificates without the dashboard. Check a site in the browser before touching anything; if the sites are up, you are fixing a control-plane problem, not an outage.

I lost the Dokploy admin password; is that why I cannot log in?

A password problem shows the login page, not a connection error. Dokploy has a documented reset procedure (Reset Password and 2FA in the docs) run from the server over SSH. If the page itself does not load, work through the six causes above first.

Why does the dashboard stop after every reboot?

Most often the start-order race: the dokploy service starts before dokploy-postgres and cannot connect. Scaling the service to 0 and back to 1 fixes it; a full disk is the other common reboot-time cause. If it recurs on every boot, check that Docker starts cleanly (journalctl -u docker) and that the disk has headroom.