ERT Storage Server

This is the permanent storage solution for the ERT project. It is written in Python 3.6+ using FastAPI, Pydantic, SQLAlchemy, Graphene and Azure Blob Storage.

Development Environment

This section describes how to use ERT Storage for testing or development purposes. It uses an in-memory SQLite database with no Azure Blob Storage.

First, install ERT Storage into a Python virtual environment:

# Clone this repo
git clone https://github.com/equinor/ert-storage
cd ert-storage

# Create the virtual environment (assuming you have Python 3.6+)
python3 -m venv my-venv

# Enable the environment
source my-venv/bin/activate       # sh/bash/zsh

# Install ERT Storage into the environment from current working directory
pip install -e .

Then start ERT Storage simply by using the new ert-storage command. This command will tell you that you are using an in-memory SQLite database. Now, using your browser, navigate to http://localhost:8000 to bring up the Swagger documentation page. Pages of note are:

• http://localhost:8000/doc Swagger
• http://localhost:8000/redoc ReDoc (alternative to Swagger)
• http://localhost:8000/gql Graph_i_QL, interactive GraphQL page

Persisting data

Specifying the ERT_STORAGE_DATABASE_URL environment variable it is possible to specify a non-in-memory database. Refer to SQLAlchemy Documentation for all possible database URLs. However, only SQLite and PostgreSQL with default SQLAlchemy drivers are officially supported at this time.

SQLite

To persist a SQLite database (note the number of forward slashes /):

export ERT_STORAGE_DATABASE_URL="sqlite:///ert.db"       # Use ert.db in current working directory
export ERT_STORAGE_DATABASE_URL="sqlite:////tmp/ert.db"  # Use /tmp/ert.db

NOTE: SQLite does not support the same features as PostgreSQL, and as such migrations (via alembic) are not supported and some data may be stored or fetched in an inefficient manner. You will have to delete your data whenever we update the database schema, which happens relatively often.

PostgreSQL

We use PostgreSQL in production and we use alembic for migrations. The following will set everything up.

Set ERT_STORAGE_DATABASE_URL

# Example postgres settings
HOST="localhost"
PORT=5432  # default PostgreSQL port
DATA="ert_store"
USER="my-usrnam"
PASS="my-passwd"

# The connection string is then:
export ERT_STORAGE_DATABASE_URL="postgresql://${USER}:${PASS}@${HOST}:${PORT}/${DATA}"

# or
export ERT_STORAGE_DATABASE_URL="postgresql://my-usrnam:my-passwd@localhost:5432/ert_store"

Then install a compatible SQLAlchemy PostgreSQL driver, such as psycopg2. We provide the extras requirement postgres which installs the necessary requirements.

pip install -e .[postgres]

The psycopg2 package requires the postgres client development headers to be installed on your machine. The psycopg2-binary is a precompiled version of this package that might be compatible with your OS and PostgreSQL version.

Database setup: Container

You can use either Docker or podman to easily start a postgres instance as a container. On modern Linux distributions it is easier to use the podman project.

See the docker.io page for postgres for more information

The following instructions are for podman. Replacing podman with docker will

# Start the container in the background, with password 'password', forwarding the port 5432 to the host.
podman run --name ert-storage-pg -e POSTGRES_PASSWORD=password -p 5432:5432 -d docker.io/postgres

# Useful commands
podman stop ert-storage-pg
podman start ert-storage-pg
podman kill ert-storage-pg  # SIGTERM
podman rm ert-storage-pg    # Remove the container entirely

# Use with ERT Storage
export ERT_STORAGE_DATABASE_URL="postgresql://postgres:password@localhost:5432/postgres"

To reset the database, do:

podman exec ert-storage-pg sh -c "dropdb -U postgres postgres; createdb -U postgres postgres"
ert-storage alembic upgrade head

Database setup: Linux service

First, install the dependencies:

# RHEL / CentOS / Fedora
sudo yum install postgresql-server postgresql-devel

# Ubuntu
sudo apt install postgresql postgresql-client

Then, initialize the database and configure a user with a password.

# Initialise stuff in /var (OPTIONAL)
sudo -i -u postgres -- initdb

# Create superuser
sudo -i -u postgres -- createuser -s ert_user

# Give yourself a password (default is to disallow password login)
sudo -i -u postgres -- psql -c "alter user ert_user with password '12345';"

# Create a database with you as its owner
createdb -U ert_user ert_db

# Set the database URL so that ERT Storage can connect to it
# You will need to redo this every time you close your shell
export ERT_STORAGE_DATABASE_URL="postgresql://ert_user:12345@localhost:5432/ert_db"

To recreate the database, do:

dropdb ert_db && createdb -O ert_user ert_db
ert-storage alembic upgrade head

After setup

Once a PostgreSQL server is running, you need to apply alembic migrations. ert-storage provides a wrapper around the alembic command which automatically selects the correct alembic configuration.

# Upgrade to latest migration
ert-storage alembic upgrade head

Azure Blob Storage

ERT Storage supports Azure Blob Storage for storing opaque data. This feature is invisible to the user. Install the azure extras with pip install ert-storage[azure].

Set the ERT_STORAGE_AZURE_CONNECTION_STRING environment variable to the connection string found in your Azure configuration to have ERT Storage use Azure Blob Storage.

Local instance for development

It is possible to use Azurite, an open source Azure Storage implementation, for testing and development purposes.

As a container:

# Podman
podman run --name ert-storage-az -p 10000:10000 -p 10001:10001 -d mcr.microsoft.com/azure-storage/azurite

# Let ERT Storage know about Azurite
export ERT_STORAGE_AZURE_CONNECTION_STRING="DefaultEndpointsProtocol=http;AccountName=devstoreaccount1;AccountKey=Eby8vdM02xNOcqFlqUwJPLlmEtlCDXJ1OUzFT50uSRZ6IFsuFq2UVErCz4I6tq/K1SZFPTOtr/KBHBeksoGMGw==;BlobEndpoint=http://127.0.0.1:10000/devstoreaccount1;QueueEndpoint=http://127.0.0.1:10001/devstoreaccount1;"

Development

For development, install the test extras pip install ert-storage[test], which installs black, pytest and mypy. Run tests using pytest.

All integration tests must be able to pass when using a non-empty production database. In practice that means that each test needs to create a unique "experiment".

During test debugging it might be helpful to have the tests persist data in the database. The following environment variable tells ERT Storage not to rollback the data after each test.

export ERT_STORAGE_NO_ROLLBACK=1