@bitscheme/mgdb-migrator v5.0.1

A simple migration system for mongodb supporting up/downwards migrations.

Status

Branch	Status
Next
Master

Install

$ npm i mgdb-migrator

or

$ yarn add mgdb-migrator

Quick Start

import { migrator } from 'mgdb-migrator'

await migrator.config({
  // false disables logging
  log: true,
  // optional logging function
  logger: (level, ...args) => console.log(level, ...args),
  // migrations collection name defaults to 'migrations'
  collectionName: 'migrations',
  // max time allowed in ms for a migration to finish, default Number.POSITIVE_INFINITY
  timeout: 30000,
  // connection properties object
  db: {
    // mongodb connection url
    connectionUrl: 'mongodb://localhost:27017/my-db',
    // optional database name, in case using it in connection string is not an option
    name: 'my-db',
    // optional mongodb MongoClientOptions
    options: {
      useNewUrlParser: true,
      useUnifiedTopology: true,
    },
  },
})

migrator.add({
  version: 1,
  name: 'Name for this migration',
  up: async (client: MongoClient, logger: Logger) => {
    // write your migration here
    await client
      .db()
      .collection('albums')
      .updateMany({}, { $set: { stars: 5 } })
  },
  down: async (client: MongoClient, logger: Logger) => {
    // write your reverting migration here
    await client
      .db()
      .collection('albums')
      .updateMany({}, { $set: { stars: 0 } })
  },
})

// run all configured migrations greater than the current version in order
await migrator.up()

Versioning

Migration versions use sequential integers. Version 0 is reserved by migrator for initial state to indicate no migrations have been applied.

Flow

Migration state is implemented in the MongoDB collection migrations. It contains a single document used for locking migration control. Only one set of migrations is allowed to execute at a time.

You can override the collection name in config if needed.

{
  _id: String, // 'control'
  version: Int32,
  locked: Bool,
  lockedAt: Date
}

When a migration is performed, all migrations that include versions between current and target are executed serially in order.

For example, if you have added the following migrations:

• v1
• v2
• v3
• v4

and the current version is at v0, executing up(3) will run migrations v1, v2 and v3. If all migrations were successful, the current version becomes v3.

If any particular migration rejects or throws an error, subsequent migrations are halted and the current version is set to the last successfully completed migration.

API

config(opts: IMigrationOptions) ⇒ Promise<void>

See the Quick Start for options.

add(migration: IMigration)

To setup a new database migration script, call migrator.add.

You must implement up and down functions. Return a promise (or use async/await) and resolve to indicate success, throw an error or reject to abort.

up(target?: number) ⇒ Promise<void>

To migrate to the latest configured migration:

migrator.up()

Or by specifying a target version, you can migrate directly to that version (if possible).

migrator.up(1)

down(target: number) ⇒ Promise<void>

To revert a migration:

migrator.down(1)

If you want to undo all of your migrations, you can migrate back down to version 0 by running:

migrator.down(0)

Sometimes (usually when something goes awry), you may need to retry a migration. You can do this by updating the migrations.version field in mongodb to the previous version and re-executing your migration.

getVersion() ⇒ number

To see what version the database is at, call:

migrator.getVersion()

getMigrations() ⇒ IMigration[]

To see the configured migrations (excluding v0), call:

migrator.getMigrations()

close(force?: boolean) ⇒ Promise<void>

To close the mongodb connection, call:

migrator.close()

Using MongoDB Transactions API

You can make use of the MongoDB Transaction API in your migration scripts.

Note: this requires

• MongoDB 4.0 or higher

migrator will call your migration up and down function with a second argument: client, a MongoClient instance to give you access to the startSession function.

Example:

const albumMigration = {
  version: 1,
  async up(client) {
    const session = client.startSession()
    try {
      await session.withTransaction(async () => {
        await db
          .collection('albums')
          .updateOne({ artist: 'The Beatles' }, { $set: { blacklisted: true } })
        await db.collection('albums').updateOne({ artist: 'The Doors' }, { $set: { stars: 5 } })
      })
    } finally {
      await session.endSession()
    }
  },
  async down(client) {
    const session = client.startSession()
    try {
      await session.withTransaction(async () => {
        await db
          .collection('albums')
          .updateOne({ artist: 'The Beatles' }, { $set: { blacklisted: false } })
        await db.collection('albums').updateOne({ artist: 'The Doors' }, { $set: { stars: 0 } })
      })
    } finally {
      await session.endSession()
    }
  },
}

Logging

Migrations uses the console by default for logging if not provided. If you want to use your own logger (for sending to other consumers or similar) you can do so by configuring the logger option when calling migrator.config.

Log levels conform to those in node.js Console API.

Winston example

import { createLogger } from 'winston';

const logger = createLogger({
  transports: [
    new winston.transports.Console();
  ]
});

const myLogger = (level, message) => {
  logger.log({
    level,
    message
  });
}

migrator.config({
  ...
  logger: myLogger
  ...
});

Development

Run docker-compose to execute lib in dev mode

$ npm run docker:dev

Test

To configure a specific db connection for testing, create a .env file in the root of the project:

Example:

MONGODB_URL=mongodb://localhost:27017/mgdb-migrator-test

Run docker-compose to execute lib in test mode

$ npm run docker:test

Credits

Migration builds on percolatestudio/meteor-migrations with the goal of creating a generic mongodb migration library