Prisma is an Object Relation Mapping (ORM) solution for TS/JS, it is by far my favorite ORM with its ability to introspect into existing databases, intuitive schema models, type-safe API and amazing tooling. With Prisma 3, the library now supports NoSQL databases like MongoDB, making it even more enticing to use for projects where relational databases are not necessary.

While the integration is amazingly smooth when you are working with a MongoDB deployment with replica sets enabled, if you try to run prisma against a database without it you may experience errors that looks something like this:

PrismaClientUnknownRequestError2 [PrismaClientUnknownRequestError]:
Invalid `prisma.post.create()` invocation in /index.ts:9:21
   6 await prisma.$connect()
   7
   8 // Create the first post
→  9 await prisma.post.create(
  Error in connector: Database error. error code: unknown, error message: Transactions are not supported by this deployment
    at cb (/node_modules/@prisma/client/runtime/index.js:34804:17)
    at processTicksAndRejections (internal/process/task_queues.js:97:5) {
  clientVersion: '3.xx.0'
}

The Prisma team explains that a replica set is required for transactions in MongoDB and that to prevent partial writes nested queries require transactions to be enabled. A simple solution to get a MongoDB instance with replica set out of the box is to deploy a free MongoDB Atlas instance, but that might not be viable some times especially if you just want a offline capable development environment.

So this guide will provide a few strategies for running a local deployment of MongoDB on Docker that will work with Prisma.

Use an image with replica sets enabled

This is the simplest solution as there are quite a few images on DockerHub that has replica sets enabled by default, the one we are using here is published by Bitnami, tagged bitnami/mongodb.

To use the image in Docker Compose, simply replace the mongo in your compose file with bitnami/mongodb:latest, and add a few environment variables needed for replica sets, for example:

version: '3.8'
services:
  database:
    image: 'bitnami/mongodb:latest'
    environment:
      - MONGODB_ADVERTISED_HOSTNAME=database
      - MONGODB_REPLICA_SET_MODE=primary
      - MONGODB_ROOT_USER=anyuser
      - MONGODB_ROOT_PASSWORD=anypassword
      - MONGODB_REPLICA_SET_KEY=replicasetkey123
    ports:
      - '27017:27017'

Note that the MONGODB_REPLICA_SET_MODE, MONGODB_REPLICA_SET_KEY are required to enable replica sets, while MONGODB_ADVERTISED_HOSTNAME is strongly recommended. Here is an explaination of what each of the environment variable does:

• MONGODB_REPLICA_SET_MODE: select the mode to run the replica set as, possible values are primary, secondary, and arbiter. For our case, primary is recommended.
• MONGODB_REPLICA_SET_KEY: the replica set key, should be longer than 5 characters.
• MONGODB_ADVERTISED_HOSTNAME: the advertised hostname, this is not required but if you find there to be an issue without it, set the value to either 127.0.0.1, host.docker.internal, localhost, or the service name depending on your setup.

This solution has a few gotchas:

1. The latest version of the image is not compatible with Apple Silicon devices, so if you know the service may need to run on such device, use bitnami/mongodb:4.4 instead.
2. The image does not have any arm optimized tags, so expect worse battery life performance if used on Apple Silicon.

Build your own image on top of mongo

If you don't want to leverage an image built by a 3rd party (albite a verified one), you can build your own replica set enabled docker image for your local development, this is a Dockerfile I found to work well, taken from Prisma's examples

FROM mongo:5

# we take over the default & start mongo in replica set mode in a background task
ENTRYPOINT mongod --port $MONGO_REPLICA_PORT --replSet rs0 --bind_ip 0.0.0.0 & MONGOD_PID=$!; \
  # we prepare the replica set with a single node and prepare the root user config
  INIT_REPL_CMD="rs.initiate({ _id: 'rs0', members: [{ _id: 0, host: '$MONGO_REPLICA_HOST:$MONGO_REPLICA_PORT' }] })"; \
  INIT_USER_CMD="db.createUser({ user: '$MONGO_INITDB_ROOT_USERNAME', pwd: '$MONGO_INITDB_ROOT_PASSWORD', roles: [ 'root' ] })"; \
  # we wait for the replica set to be ready and then submit the commands just above
  until (mongo admin --port $MONGO_REPLICA_PORT --eval "$INIT_REPL_CMD && $INIT_USER_CMD"); do sleep 1; done; \
  # we are done but we keep the container by waiting on signals from the mongo task
  echo "REPLICA SET ONLINE"; wait $MONGOD_PID;

Then add the image and necessary environment variables to your compose yaml:

version: '3.8'

services:
  database:
    build: ./database-img
    environment:
      - MONGO_INITDB_ROOT_USERNAME=anyuser
	  - MONGO_INITDB_ROOT_PASSWORD=anypassword
	  - MONGO_INITDB_DATABASE=database
	  - MONGO_REPLICA_HOST=172.17.0.1
	  - MONGO_REPLICA_PORT=27017
    ports:
      - 27017:27017

For this image, all 5 environment variables are necessary as they are used for initializing the database and creating the root user. It is recommended that you place the Dockerfile in a separate directory, they way I structured it is:

project
│   docker-compose.yml
└───database-img
│   │   Dockerfile

That's it! You can test the connection with your database by using prisma db pull or prisma db push