This docker image provides a method to run migrations and metadata at docker entrypoint. A temporary server is booted, with the migrations API allowed, securely through localhost. Once migrations and metadata have been applied, the server will reboot in a secure mode for inbound graphql usage.
It is used in a docker file as:
FROM hasura/graphql-engine:<version>.cli-migrations-v2
CMD graphql-engine \
--database-url $DATABASE_URL \
serve \
--server-port $PORT \
--enable-console
This is covered in the documentation here: https://hasura.io/docs/latest/graphql/core/migrations/auto-apply-migrations.html The below Configuration will also be applicable.
Using a docker build on heroku the manifest (heroku.yml
) can look like:
setup:
addons:
- plan: heroku-postgresql
as: DATABASE
config:
HASURA_GRAPHQL_MIGRATIONS_DATABASE_ENV_VAR: DATABASE_URL
build:
docker:
web: Dockerfile
This allows you to provision and migrate a database entirely without intervention.
Setup the app, with addons and the setup steps defined in the heroku.yml
heroku create heroku-migration-tester --manifest
export HEROKU_GIT_REMOTE=https://git.heroku.com/heroku-migration-tester.git
git init && git add .
git commit -m "first commit"
git remote add heroku HEROKU_GIT_REMOTE
git push heroku master
You may set the following environment variables to configure the way this image is run.
Migrations are either mounted or built into the image.
If it has been stored in a directory other than the default then it can be configured using the following:
-
HASURA_GRAPHQL_MIGRATIONS_DIR
(default=/hasura-migrations
)A path to the migrations directory.
Metadata are either mounted or built into the image.
If it has been stored in a directory other than the default then it can be configured using the following:
-
HASURA_GRAPHQL_METADATA_DIR
(default=/hasura-metadata
)A path to the metadata directory.
The following are listed in order of evaluation, therefore if setting the first, the latter will not be evaluated. At least one of these must be configured to migrate the database successfully.
-
HASURA_GRAPHQL_MIGRATIONS_DATABASE_ENV_VAR
(default=null
)Defines a pointer to an environment variable, which holds the database url e.g.
HASURA_GRAPHQL_MIGRATIONS_DATABASE_ENV_VAR=DATABASE_URL
-
HASURA_GRAPHQL_MIGRATIONS_DATABASE_URL
(default=null
)Defines the database url for migrations, allowing it to be separate from the database used for querying. An example use case is for a read only follower database used for querying your graphql endpoint. In this scenario you would migrate only the master and the follower would be kept in sync.
HASURA_GRAPHQL_MIGRATIONS_DATABASE_URL=postgres://<username>:<password>@<host>:<port>/<database_name>
-
HASURA_GRAPHQL_DATABASE_URL
(default=null
)Use as above to point to the default database url.
HASURA_GRAPHQL_DATABASE_URL=postgres://<username>:<password>@<host>:<port>/<database_name>
Optional configuration for the server which boots during migrations.
-
HASURA_GRAPHQL_MIGRATIONS_SERVER_PORT
(default=9691
)Specify the port running the graphql server during execution of the migration script. It is advised that you do not specify a PORT that may be open e.g. 80/443 and the default should rarely require changing.
-
HASURA_GRAPHQL_MIGRATIONS_SERVER_TIMEOUT
(default=30s
)Specify the server timeout threshold.
-
HASURA_GRAPHQL_DISALLOW_INCONSISTENT_METADATA
(default=false
)If set to true, inconsistent metadata will be rejected, causing the entrypoint script to exit with a non-zero exit code. Has the same behavior as the
--disallow-inconsistent-metadata
flag in thehasura metadata apply
command documented here.