migrado

Migrado

🥑 ArangoDB migrations and batch processing manager.

Migrado is a command-line client that can help build and run schema or data migrations against your ArangoDB instance.

Migrado utilizes ArangoDB Transactions when running data migrations to ensure failed scripts are rolled back automatically. arangosh from the ArangoDB Client Tools is required to run schema migrations, however no transaction safety is available at this point.

Migrado should be considered beta software, but it is well tested, and used in production settings. Make sure you understand how it operates.

If you have trouble, open an issue. Contributions are welcome.

Installation

Migrado requires Python 3.6 or higher, and the ArangoDB arangosh client.

$ pip install --user migrado

It is also available as a Docker image, see Docker usage.

Usage

Migrado can create a migrations directory and generate an initial set of collections from a given schema file:

$ migrado init --schema schema.yml

Migrado can also construct an initial migration from the current database structure (and automatically store it as the current state/schema):

$ migrado init --infer

See YAML schemas for details. If neither option is specified, Migrado will create an empty initial migration.

To autogenerate a schema migration script based on an updated schema:

$ migrado make --schema updated_schema.yml

To make a new template data migration script:

$ migrado make --name rewrite_field_names

This will create a new file, migrations/0002_rewrite_field_names.js (--name is optional), which you can edit as you see fit. See Migration scripts for details.

When you are ready, run all migrations (not previously ran) against the database:

$ migrado run

Migrado stores migration state in a configurable collection, see --help or Environment vars for details.

If you wrote a reverse() migration, you can revert to an earlier point by specifying a target migration id. To revert to the initial migration:

$ migrado run --target 0001

You can inspect the current migration state with:

$ migrado inspect

You can inspect the current schema (explicit or inferred) with:

$ migrado export

Use the --help option for help on any command when using the client.

Docker usage

If you're using Migrado in a Docker context, you might as well use the Docker image. migrado is set as entrypoint, so the image can be used like the Python client:

$ docker run protojour/migrado --help

You'd want to volume in your migrations folder:

$ docker run -v /path/to/migrations:/app/migrations protojour/migrado

Or, an example using docker-compose:

migrado:
  image: protojour/migrado:latest
  environment:
    MIGRADO_DB: # ...
    MIGRADO_HOST: # ...
  volumes:
    - ./migrations:/app/migrations

Then either add a command: (with a migrado sub-command, e.g. command: run ...), or use this as a starting point for a scripted migration strategy.

You may also use the base Dockerfile as a starting point.

Environment vars

The following environment variables are employed by Migrado:

• MIGRADO_PATH: Specifies the path to the migrations directory, replaces -p, --path (default: migrations).
• MIGRADO_DB: Specifies the ArangoDB database name for generated migrations to interact with, replaces -d, --db (no default, but required for the run command).
• MIGRADO_COLL: Specifies ArangoDb collection name to store migration state in, replaces -c, --state-coll (default: migrado).
• MIGRADO_TLS: Use TLS for connection when running migrations, replaces -T, --tls (default: False).
• MIGRADO_HOST: Specifies the database host for running migrations, replaces -H, --host (default: localhost).
• MIGRADO_PORT: Specifies the database port for running migrations, replaces -P, --port (default: 8529).
• MIGRADO_USER: Specifies the database username for running migrations, replaces -U, --username (no default).
• MIGRADO_PASS: Specifies the database password for running migrations, replaces -W, --password (no default).

YAML schemas

ArangoDB may be schemaless, but in a larger project it still makes sense to keep a schema spec up to date, both for an overview of collections and their data structures, and as a basis for native collection-level validation (see the -v/--validation option).

Migrado uses a schema model based on JSON Schema, in YAML, and can use this to generate an initial migration for the collections available in your database.

Example schema:

all: &all
  _id:
    type: string
    readOnly: true
  _key:
    type: string
    readOnly: true
  _rev:
    type: string
    readOnly: true

edges: &edges
  _from:
    type: string
  _to:
    type: string

collections:

  books:
    type: object
    properties:
      <<: *all
      title:
        type: string
      isbn:
        type: string
    required:
      - title
      - isbn

  authors:
    # Note: you do not actually need to specify an object schema,
    # but the schema can be used with ArangoDB's native validation 
    # using the -v/--validation option

edge_collections:

  # authors --> books 
  author_of:
    type: object
    properties:
      <<: *all
      <<: *edges
    required:
      - _from
      - _to

Migration scripts

Migration scripts are structured so they may be parsed and run easily by both Migrado and ArangoDB. In addition, they are structured so they may be run manually against ArangoDB using arangosh.

There are two types of script, data and schema migration scripts.

Data migrations

You need to declare all collections subject to write operations using the syntax // write collection_name, because ArangoDB needs this information for locking during transactions. We've made the declaration explicit to reduce errors. Attempting to write to collections not declared in this way will cause the migration to fail.

In general, a reverse migration should do the logical opposite of a forward migration. forward() and reverse() functions can contain anything that the ArangoDB V8 engine understands, but must be fully self-contained. Anything outside these functions is ignored and unavailable when running migrations.

Here's an example migration script for adding new_field in collection things:

// write things

function forward() {
    var db = require("@arangodb").db
    db._query(`
        FOR thing IN things
            UPDATE thing WITH { new_field: "some value" } IN things
    `)
}

function reverse() {
    var db = require("@arangodb").db
    db._query(`
        FOR thing IN things
            REPLACE thing WITH UNSET(thing, "new_field") IN things
    `)
}

Please make sure you read limitations when running transactions in the ArangoDB documentation. In particular, creation and deletion of databases, collections, and indexes is not allowed in transactions.

Schema migrations

Schema migrations are stuctured in the same way as data migrations, but are run against arangosh as opposed to the HTTP API. There is no transaction safety when running schema migrations.

Schema migrations are structured the same way as data migrations, but // write declarations are not required. All operations are allowed.

Here's an example migration script generated from the YAML schema above (with no validation):

function forward() {
    var db = require("@arangodb").db
    var schema = // schema to be written to disk
    db._create("books", {}, "document")
    db._create("authors", {}, "document")
    db._create("author_of", {}, "edge")
}

function reverse() {
    var db = require("@arangodb").db
    db._drop("books")
    db._drop("authors")
    db._drop("author_of")
}

Please be careful when running schema migrations in reverse. As you can see, the reverse() function above would drop your collections (and lose your data) if you were to reverse beyond this point. Currently, you will not be able to do so for an initial migration.

License

Migrado is copyright © 2019 Protojour AS, and is licensed under MIT. See LICENSE.txt for details.