Setting up Hasura migrations

Introduction

If you don’t already use any tool to manage your Postgres schema, you can use Hasura to do that for you. Hasura has a CLI which will help you save each action that you do on the console, including creating tables/views and schema modifying SQL statements, as SQL files. These files are called migrations and they can be applied and rolled back step-by-step. These files can be version controlled and can be used with your CI/CD system to make incremental updates.

Let's say we have the following two tables in our schema:

author (id uuid, name text, rating integer)
article (id uuid, title text, content text, author_id uuid)

Now we want to set up migrations starting with this schema.

Step 0: Disable the console served by the server

The Hasura console UI can be served directly by the GraphQL engine server and also by the Hasura CLI.

To use migrations effectively, the console served by the server (which is served at /console) should be disabled and all changes must go through the console served by the CLI. Otherwise, changes could be made through the server console and they will not be tracked by migrations.

So, the first step is to disable the console served by the GraphQL engine server. In order to do that, remove the --enable-console flag from the command that starts the server or set the following environment variable to false:

HASURA_GRAPHQL_ENABLE_CONSOLE=false

Note

If this is set in YAML, make sure you quote the word false, i.e. HASURA_GRAPHQL_ENABLE_CONSOLE: "false".

Step 1: Install the Hasura CLI

Follow the instructions in Installing the Hasura CLI.

Step 2: Set up a project directory

For the endpoint referred here, let's say you've deployed the GraphQL engine on Hasura Cloud, then this endpoint is: https://my-graphql.hasura.app. In case you've deployed Hasura using Docker, the URL might be http://xx.xx.xx.xx:8080. In any case, the endpoint should not contain the v1/graphql API path. It should just be the hostname and any sub-path if it is configured that way.

Let's set up a project directory by executing the following command:

hasura init my-project --endpoint http://my-graphql.hasura.app

cd my-project

This will create a new directory called my-project with a config.yaml file, a migrations directory and a metadata directory. This directory structure is mandatory to use Hasura migrations.

note

In case there is an admin secret set, you can set it as an environment variable HASURA_GRAPHQL_ADMIN_SECRET=<your-admin-secret> on your local machine and the CLI will use it. You can also use it as a flag to CLI commands: --admin-secret '<your-admin-secret>'.

Step 2.1: Set up version control for your project directory

The project directory created above can be committed to version control.

Set up version control and commit the project status:

# in project dir

# initialize version control
git init

# commit initial project status
git add .
git commit -m "hasura project init"

Step 3: Initialize the migrations and metadata as per your current state

If you have already set up your database and GraphQL API, you need to initialize your database migrations and Hasura metadata with the current state of the database.

Step 3.1: Initialize database migrations

Create a migration called init by exporting the current Postgres schema from the server:

# create migration files (note that this will only export the public schema from postgres)
# if you started hasura with HASURA_GRAPHQL_DATABASE_URL  environment variable, the database name should be default
hasura migrate create "init" --from-server --database-name <database-name>

# note down the version

# mark the migration as applied on this server
hasura migrate apply --version "<version>" --skip-execution --database-name <database-name>

This command will create a new directory named <timestamp>_init inside the migrations/<database-name> directory. In the newly created directory, there's a file named up.sql. This file will contain the required information to reproduce the current state of the server including the Postgres (public) schema. If you'd like to read more about the format of migration files, check out the Migration file format reference.

The apply command will mark this migration as "applied" on the server.

Note

If you need to export other schemas along with public, you can name them using the --schema flag.

For example, to export schemas public, schema1 and schema2, execute the following command:

hasura migrate create "init" --from-server --schema "public" --schema "schema1" --schema "schema2" --database-name <database-name>

Step 3.2: Initialize Hasura metadata

Export the Hasura metadata from the server:

# export the metadata
hasura metadata export

This command will export the current Hasura metadata as a bunch of YAML files in the metadata directory.

If you'd like to read more about the format of metadata files, check out the Metadata format reference.

Step 3.3: Add a checkpoint to version control

Commit the current project state to version control:

# in project dir
git add .
git commit -m "initialize migrations and metadata"

note

The version control set up should typically be done right after Step 2

Step 4: Use the console served by the CLI

From this point onwards, instead of using the console at http://my-graphql.hasura.app/console you should use the console served by the CLI by running:

# in project dir
hasura console

Step 5: Add a new table and see how migrations and metadata is updated

As you use the Hasura console UI served by the CLI to make changes to your schema, database migration files are automatically generated in the migrations/ directory and the metadata is exported in the metadata/ directory of your project.

Let's create the following table address (id uuid, street text, zip text, city text, country text, author_id uuid) and then create a foreign-key to the author table via the author_id -> id columns.

In the migrations/<database-name> directory, we can find new directories called <timestamp>_create_table_public_address and <timestamp>_set_fk_public_address_author_id containing an up.sql and a down.sql migration files for the changes we made.

You can also go ahead and add permissions and create relationships for the address table. The related metadata changes will automatically be exported into the metadata directory.

Note

Migrations are only created when using the console through the CLI.

Step 6: Squash migrations and add checkpoints to version control

As you keep using the console via the CLI to make changes to the schema, new migration files will keep getting generated and the metadata files will keep getting updated automatically.

Typically while adding a feature a lot of incremental migration files get created for each of the small tasks that you did to achieve the feature. To improve maintainability of the migration files and to ensure you can go back to a particular version of the metadata, it is recommended that you squash your migration files and commit the project status in version control whenever you reach a logical checkpoint in your feature development.

The following command will squash all migration files from the given migration to the latest migration into a single migration file.

hasura migrate squash --name "<feature-name>" --from <start-migration-version> --database-name <database-name>

# note down the version

# mark the squashed migration as applied on this server
hasura migrate apply --version "<squash-migration-version>" --skip-execution --database-name <database-name>

Commit the project status into version control.

# in project dir
git add .
git commit -m "<feature-name>"

Note

The version control set up should typically be done right after Step 2

Step 7: Apply the migrations and metadata on another instance of the GraphQL engine

Apply all migrations present in the migrations/ directory and the metadata present in the metadata/ directory on a new instance at http://another-graphql-instance.hasura.app:

# in project dir

# apply metadata, this will connect Hasura to the configured databases.
hasura metadata apply --endpoint http://another-graphql-instance.hasura.app
# apply migrations to the connected databases.
hasura migrate apply --all-databases --endpoint http://another-graphql-instance.hasura.app
# reload metadata to make sure Hasura is aware of any newly created database objects.
hasura metadata reload --endpoint http://another-graphql-instance.hasura.app

In case you need an automated way of applying the migrations and metadata, take a look at the cli-migrations Docker image, which can start the GraphQL engine after automatically applying the migrations and metadata which are mounted onto directories.

As we are applying metadata before migrations, your metadata might be in an inconsistent state after the initial metadata apply till the metadata reload step as some database objects referred to in metadata might not be available till the migrations are applied.

If you now open the console of the new instance, you can see that the three tables have been created and are tracked:

Step 8: Check the status of migrations

# in project dir
hasura migrate status --database-name <database-name>

This command will print out each migration version present in the migrations directory along with its name, source status and database status.

For example,

$ hasura migrate status --database-name <database-name>
VERSION        NAME                           SOURCE STATUS  DATABASE STATUS
1590493510167  init                           Present        Present
1590497881360  create_table_public_address    Present        Present

Such a migration status indicates that there are 2 migration versions in the local directory and both of them are applied on the database.

If SOURCE STATUS indicates Not Present, it means that the migration version is present on the server, but not on the current user's local directory. This typically happens if multiple people are collaborating on a project and one of the collaborators forgot to pull the latest changes which included the latest migration files, or another collaborator forgot to push the latest migration files that were applied on the database. Syncing of the files would fix the issue.

If DATABASE STATUS indicates Not Present, it denotes that there are new migration versions in the local directory which are not applied on the database yet. Executing hasura migrate apply will resolve this.