Install with Docker

หมายเหตุ: บทความนี้ยังเป็นภาษาอังกฤษ เนื่องจากคำแปลภาษาไทยกำลังอยู่ระหว่างจัดทำ

Purpose

Run immutable EmuFramework images with persistent database storage.

Audience

Docker operators and framework administrators.

Prerequisites

Docker Engine with Compose and access to ghcr.io.

Procedure

Obtain docker-compose.yml from the framework repository. Cloning the repository is not required.
Create .env beside docker-compose.yml:
```
EMU_VERSION=0.1.1.0
EMU_UPDATER_TOKEN=replace-with-a-random-secret-at-least-24-characters
PORT=3399
EMU_SECURE_COOKIES=true
```
The token is created by the operator; it is not downloaded from GitHub. Use the same token for the app and updater, and never commit .env.

Run:

docker compose pull
docker compose up -d
docker compose ps

Read the one-time administrator setup code:
Terminal window
```
docker compose logs app
```
Open http://localhost:3399, enter the code, choose the administrator username, and set a password of at least 12 characters.

The code expires after 15 minutes or ten failed attempts. Restart the app container to generate a new code:

docker compose restart app
docker compose logs --tail=100 app

The Compose file supplies the app with EMU_DEPLOYMENT_MODE=docker, connects it to the internal updater at http://updater:3400, and mounts the named emu-data volume at /data.

The default integration secret key is /data/.emu-secret.key, so it persists in the same named volume. It is not included in .emubackup exports; keep a separate secure copy after configuring SMTP. Set EMU_SECRET_KEY_PATH only when the replacement path is also mounted persistently.

Overriding the volume and network names

Compose already gives the volume and network explicit names, so backup, monitoring, migration, and external service connections can reference them reliably. To use different names — for example to avoid a collision with another deployment on the same host — set these before the first docker compose up:

EMU_VOLUME_NAME=mycompany-emu-data
EMU_NETWORK_NAME=mycompany-emu-network

Changing these after the volume and network already exist does not rename them; set them before first creating the stack, or migrate data to the new volume as described below.

Image source

Both images are published to the GitHub Container Registry (ghcr.io) under the emu479p01 org, listed at Your Packages:

ghcr.io/emu479p01/emu-framework:<version>          # app
ghcr.io/emu479p01/emu-framework-updater:<version>  # updater

docker-compose.yml resolves <version> from EMU_VERSION in .env, so docker compose pull fetches both images at that tag. No account or authentication is required to pull, since the packages are public. Pulling manually without Compose looks like:

docker pull ghcr.io/emu479p01/emu-framework:<version>
docker pull ghcr.io/emu479p01/emu-framework-updater:<version>

To use a specific version instead of the latest, set EMU_VERSION accordingly before running docker compose pull.

Docker Desktop without cloning

Pulling an image alone is not a complete installation. docker pull downloads an image but does not create the app, updater, network, environment variables, or persistent volume.

Create the shared network and volume first:

docker network create emu-network
docker volume create emu-data
docker network inspect emu-network
docker volume inspect emu-data

If Docker reports that the network or volume already exists, it can be reused, but verify it belongs to this EmuFramework installation before continuing.

The app container must have:

Image: ghcr.io/emu479p01/emu-framework:<version>
Container name: emu-framework
Restart policy: Unless stopped
Port: 3399 -> 3399
Network: emu-network
Volume: emu-data -> /data (Read/Write)
NODE_ENV=production
EMU_APP_TITLE=EmuFramework
EMU_DEPLOYMENT_MODE=docker
EMU_UPDATER_URL=http://updater:3400
EMU_UPDATER_TOKEN=<same token as updater>
EMU_DB_PATH=/data/data.db
EMU_DESIGNER_DB_PATH=/data/designer.db
EMU_SECRET_KEY_PATH=/data/.emu-secret.key

After starting a manually created app container, run docker logs --tail 100 emu-framework, copy the one-time setup code, and complete Administrator setup at http://localhost:3399. Restart the container if the code expires.

The updater container must use the matching updater image, the same token, the same emu-network network, the emu-data volume, and a Docker socket mount:

Image: ghcr.io/emu479p01/emu-framework-updater:<version>
Container name: emu-framework-updater
Restart policy: Unless stopped
Network: emu-network
Network alias: updater
Volume: emu-data -> /data (Read/Write)
Docker socket: /var/run/docker.sock -> /var/run/docker.sock (Read/Write)
EMU_UPDATER_TOKEN=<same token as app>
EMU_APP_CONTAINER=emu-framework
EMU_IMAGE_REPOSITORY=ghcr.io/emu479p01/emu-framework
EMU_UPDATE_STATE_PATH=/data/update-status.json

Do not publish updater port 3400 to the host; it is an internal service. The updater’s container name can be anything, but it must carry the network alias updater, because the app reaches it at http://updater:3400. If the containers were created without a network, connect them afterward:

docker network connect emu-network emu-framework
docker network connect --alias updater emu-network emu-framework-updater

After changing environment variables, recreate the app container. Restarting an existing container does not add new environment variables.

Verify app-to-updater connectivity

docker exec emu-framework node -e "fetch('http://updater:3400/').then(r=>console.log(r.status)).catch(e=>{console.error(e);process.exit(1)})"

An HTTP 404 response confirms the network and DNS alias work; the updater has no GET / route, but the app reached the service successfully. Check logs for either side if this fails:

docker logs --tail 100 emu-framework
docker logs --tail 100 emu-framework-updater

Existing installation and data migration

Before removing an existing container, inspect its mounts:

docker inspect <app-container> --format '{{json .Mounts}}'

The database files should be mounted at /data. A named volume such as emu-data can be reused safely when recreating the container. Do not run docker compose down -v unless deleting all application data is intentional. Compose names its containers emuframework-app and emuframework-updater; manually created containers are commonly named emu-framework and emu-framework-updater. List actual names with:

docker ps -a --format "table {{.Names}}\t{{.Image}}"

Find the volume mounted at /data for the app container:

docker inspect emu-framework --format '{{range .Mounts}}{{if eq .Destination "/data"}}{{.Name}}{{end}}{{end}}'

A long hexadecimal result is normally an anonymous volume. Verify it before continuing:

docker volume inspect <SOURCE_VOLUME>

Stop both the app and the updater before copying, so the SQLite database and its WAL files are in a consistent state:

docker stop emu-framework emu-framework-updater
docker volume create emu-data

Copy the data. This command refuses to run if emu-data already contains files, preventing an accidental merge of two installations:

docker run --rm --mount type=volume,src=<SOURCE_VOLUME>,dst=/source,readonly --mount type=volume,src=emu-data,dst=/target alpine:3.20 sh -c 'if [ -n "$(find /target -mindepth 1 -maxdepth 1 -print -quit)" ]; then echo "ERROR: emu-data is not empty"; exit 1; fi; cp -a /source/. /target/'

Inspect the copied files before changing or removing anything:

docker run --rm --mount type=volume,src=emu-data,dst=/data,readonly alpine:3.20 sh -c "find /data -maxdepth 2 -type f -exec ls -ln {} ';'"

Expect to find data.db, designer.db, backups/, and update-status.json; the exact list depends on features previously used. Recreate both containers with emu-data mounted at /data — a running container’s volume mount cannot be changed in place:

docker compose up -d --force-recreate

For manually managed containers, remove only the stopped containers (never the source volume yet) and create them again with emu-data -> /data for both app and updater. Confirm the mount on each:

docker inspect emu-framework --format '{{range .Mounts}}{{println .Name "->" .Destination}}{{end}}'
docker inspect emu-framework-updater --format '{{range .Mounts}}{{println .Name "->" .Destination}}{{end}}'

Before removing the old volume, verify: the app opens at http://localhost:3399; existing tables and transactions are intact; reports and installed fonts still work; the backup list still appears; the app can reach the updater; and Check for updates succeeds without fetch failed. Keep the old volume until all of this is confirmed and a separate backup exists:

docker volume rm <SOURCE_VOLUME>

Separate database volumes

By default data.db and designer.db share the single emu-data volume mounted at /data. To isolate them on separate volumes:

Add a second named volume, for example emu-designer-data, to docker-compose.yml and mount it on the app container alongside the existing one:
```
emu-data:/data
emu-designer-data:/designer-data
```

Set:

EMU_DB_PATH=/data/data.db
EMU_DESIGNER_DB_PATH=/designer-data/designer.db

Before recreating the container, create a verified backup and inspect the existing mount:
Terminal window
```
docker inspect <app-container> --format '{{json .Mounts}}'
```

Create the new volume and copy designer.db (and its -wal/-shm files if present) into it:

docker volume create emu-designer-data
docker run --rm -v emu-data:/from -v emu-designer-data:/to alpine sh -c 'cp -a /from/designer.db* /to/'

Recreate the app container so it picks up the new environment variables:
Terminal window
```
docker compose up -d --force-recreate app
```
Verify data.db remains in emu-data and designer.db now lives in emu-designer-data, then confirm the app starts and reads both correctly before removing designer.db from the original volume.

Do not run docker compose down -v during this process; it deletes named volumes.

Expected result

The app uses the emu-data volume. The updater has no public port.

Common errors

Do not commit .env.
The updater mounts /var/run/docker.sock, which grants host-level Docker control. Disable the updater and use the manual procedure if that risk is unacceptable.
EMU_UPDATER_URL is normally http://updater:3400; it is a Docker service name, not a public URL and not localhost.
Deployment: unsupported means the app container did not receive EMU_DEPLOYMENT_MODE=docker; setting the variable only on the updater is insufficient.
If the app cannot resolve updater, the two containers are not on the same user-defined network.
If port 3399 is already allocated, stop the old app container or choose another host port, for example 3401:3399.

Docker operations · Framework update