Troubleshooting

The most common issues during install, with one-line fixes you can copy and paste. Expand the one that matches your situation.

❌ "Cannot connect to the Docker daemon"

Docker isn't running. Start it:

macOS / Windows: open the Docker Desktop app and wait for the whale icon to stop animating.
Linux: sudo systemctl start docker (and sudo systemctl enable docker to start it on boot).

Re-run your install command afterwards.

❌ Port 8888 is already in use

Most often this is a previous Olympus installer that is still running, not another application. Find out which before doing anything — the fix is completely different:

docker ps -a --filter publish=8888 --format "table {{.Names}}\t{{.Image}}\t{{.Status}}"

Lists olympus-setup → a previous Olympus install holds the port. Go to A.
Prints nothing → another program on your machine holds it. Go to B.

A. A previous Olympus install has the port

Pick the option that matches what you want. None of these delete downloaded Docker images, so nothing is re-downloaded afterwards.

A1 — Just reopen the wizard. If the container above shows as Up, the installer is still running and there is nothing to fix. Open it:

http://localhost:8888

A2 — Restart everything, keeping your existing deployment. Use this after a reboot, or if the wizard became unresponsive. No data is lost:

# Linux / macOS
docker start olympus-setup
cd ~/olympus/app && docker compose up -d      # only if you already deployed

# Windows PowerShell
docker start olympus-setup
cd "$env:USERPROFILE\olympus\app"; docker compose up -d

A3 — Wipe everything and do a clean fresh install.

This frees the port and gives you a clean slate. The full procedure is "I want to start over from scratch" at the bottom of this page — it is kept in one place so the commands cannot drift.

⚠️ It permanently deletes all Olympus data. Docker images are kept, so the reinstall is fast.

B. Another program has the port

Find out what it is:

# Linux / macOS
lsof -i :8888

# Windows PowerShell
Get-NetTCPConnection -LocalPort 8888 | Select-Object -ExpandProperty OwningProcess | ForEach-Object { Get-Process -Id $_ }

Then either stop that program, or run the installer on a different host port — change -p 8888:8888 to e.g. -p 9999:8888 in your install command and open http://localhost:9999 instead.

Port 8888 belongs to the installer, not to the platform. Once deployment finishes you reach Olympus on your own domain over ports 80/443, so using a different installer port has no lasting effect.

❌ "denied: requested access to the resource is denied" when pulling the image

The install command ships with a read-only Docker Hub credential, so this usually means Docker isn't passing it through. Most often this happens if you have a personal docker login overriding it. Quick test:

docker logout

Then re-run the install command. If it still fails, double-check that you copy-pasted the command exactly, including the DOCKER_USERNAME and DOCKER_TOKEN lines.

❌ The first install is very slow or appears stuck

The full platform is ~6 GB across 30+ container images and is downloaded in parallel after the wizard hits the Deploy step. On a 100 Mbps connection this takes ~10 minutes; on slow or metered connections, considerably longer.

You can watch download progress in your original terminal (where docker logs -f olympus-setup is still running) — look for lines like Pulling fs layer, Download complete, etc.

If a single image truly stalls, retry the wizard's Deploy step — it picks up where it left off; previously-downloaded layers are reused.

❌ I can't reach http://<server-ip>:8888 from another machine

Your server is reachable on the network but the port is blocked by a firewall.

Linux (ufw): sudo ufw allow 8888/tcp
Linux (firewalld): sudo firewall-cmd --add-port=8888/tcp --permanent && sudo firewall-cmd --reload
Windows: open Windows Defender Firewall → Inbound Rules → New Rule → TCP → Specific local port 8888 → Allow.
Cloud VM (AWS / GCP / Azure): open inbound TCP 8888 in your VM's security group / network firewall rules.

note

For production, after install completes you'll only need ports 80 and 443 open, not 8888.

❌ The setup container exited or I lost the logs

Re-attach to the running container's logs:

docker logs -f olympus-setup

If docker ps -a | grep olympus-setup shows the container as Exited, restart it:

docker start olympus-setup && docker logs -f olympus-setup

If it keeps exiting, capture the full log for support:

docker logs olympus-setup > olympus-setup.log 2>&1

❌ I want to start over from scratch

:::danger This permanently deletes all your Olympus data Uploaded files, user accounts, chat history and embeddings — there is no undo. Docker images are kept, so the reinstall is fast.

It also removes your downloaded AI models (app_ollama_data, app_sd_models), often 5–30 GB. To keep those, add | grep -vE "ollama_data|sd_models" before xargs on the volume line (PowerShell: | Where-Object { $_ -notmatch "ollama_data|sd_models" }). :::

Run the block for your OS in order — the installer must go before the networks, and you must leave the folder before deleting it.

# Linux / macOS
cd ~/olympus/app 2>/dev/null && docker compose down -v --remove-orphans

# The installer is a standalone container, so the label sweep below does NOT match
# it — and it stays attached to app_gateway-public, which blocks removing that
# network. It has to go first.
docker rm -f olympus-setup ollama sd-webui 2>/dev/null

# Sweep anything left over — scoped to Olympus, nothing else on your machine
docker ps -aq        --filter label=com.docker.compose.project=app | xargs -r docker rm -f
docker volume ls -q  --filter label=com.docker.compose.project=app | xargs -r docker volume rm -f
docker network ls -q --filter label=com.docker.compose.project=app | xargs -r docker network rm

# Step out of the folder before deleting it
cd ~ && rm -rf ~/olympus

# Windows PowerShell
cd "$env:USERPROFILE\olympus\app"; docker compose down -v --remove-orphans

# The installer is a standalone container, so the label sweep below does NOT match
# it — and it stays attached to app_gateway-public, which blocks removing that
# network. It has to go first.
docker rm -f olympus-setup ollama sd-webui 2>$null

docker ps -aq        --filter label=com.docker.compose.project=app | ForEach-Object { docker rm -f $_ }
docker volume ls -q  --filter label=com.docker.compose.project=app | ForEach-Object { docker volume rm -f $_ }
docker network ls -q --filter label=com.docker.compose.project=app | ForEach-Object { docker network rm $_ }

# Windows will not delete a folder your shell is sitting in — leave it first
cd $env:USERPROFILE
Remove-Item -Recurse -Force "$env:USERPROFILE\olympus"

:::caution Deleting the folder alone is not enough The data lives in Docker volumes, not in ~/olympus. Removing only the folder leaves all 11 volumes behind — and because the Compose project keeps the same name, the next install reuses the old Postgres volume, skips init.sql, and comes up with no admin account. Always run the volume sweep above. :::

Then re-run the Install command.

Still stuck?

🐛 Open an issue on GitHub
✉️ Email support@olympus.io — please include the output of docker logs olympus-setup so we can see what happened.

Still stuck?​

Still stuck?