GraphSense REST Interface

The GraphSense REST Interface provides access to denormalized views computed by the graphsense-transformation pipeline. It is used by the graphsense-dashboard component.

It is based on a server stub generated by the OpenAPI Generator project against the graphsense-openapi specification version 1.7.0. It uses the Connexion library on top of aiohttp.

Note: This README.md is also generated. Changes must be made in templates/README.mustache.

Setup

Copy the configuration file template instance/config.yaml.template

cp instance/config.yaml.template instance/config.yaml

Open instance/config.yaml and configure the database connection.

Requirements

Make sure you are running Python 3.9 or Python 3.10.

python3 --version

Create a Python environment for required dependencies

python3 -m venv venv

Activate the environment

. venv/bin/activate

Install the requirements

pip install -r requirements.txt

You need access to GraphSense raw and transformed keyspaces. See Graphsense Transformation Pipeline for further details.

Run the REST interface

python -m aiohttp.web -H localhost -P 5000 openapi_server:main

Test the service in your browser:

http://localhost:5000

Deployment

In order to utilize multiple cores, you may run the REST interface through Gunicorn.

Install gunicorn:

pip install gunicorn

If unspecified gunicorn is run with a number of workers and threads both equal to the number of CPUs x 2.

Run production server, overriding number of workers and threads through command line options (4 workers, 4 threads). Specify the aiohttp specific WebWorker class to leverage the full power of lightweight threads:

gunicorn -w 4 --threads 4 "openapi_server:main()" --worker-class aiohttp.GunicornWebWorker

Deployment with docker

To run the server on a Docker container, please execute the following from the root directory:

Build the image

docker build -t openapi_server .

You may specifiy the number of workers and threads through the environments variables NUM_WORKERS and NUM_THREADS respectively.

Start up a container with 4 workers and 4 threads:

docker run -e NUM_WORKERS=4 -e NUM_THREADS=4 -p 9000:9000 openapi_server

Customize REST interface configuration

Flask configuration parameters can be added at the top level of instance/config.yaml. A configuration template file is provided in conf/config.yaml.template.

You may define ALLOWED_ORIGINS in instance/config.yaml if the REST interface is to be consumed by a Web browser, e.g. by graphsense-dashboard.

Logging

REST log messages can be delivered via mail. See instance/config.yaml.template for an example.

Development

Generate server stub

BRANCH=master
docker run --rm \
  -v "${PWD}:/build" \
  -v "${PWD}/templates:/templates" \
  openapitools/openapi-generator-cli \
  generate -i "https://raw.githubusercontent.com/graphsense/graphsense-openapi/${BRANCH}/graphsense.yaml" \
  -g python-aiohttp -o /build -t /templates

The service implementation (gsrest/service) should use the generated models located in openapi_server/models.

Source code generation is based on templates. See the templates directory for currently used templates. If you need more templates, retrieve them from the generator like so:

docker run --rm -v "/tmp/templates:/templates" openapitools/openapi-generator-cli author template -g python-aiohttp -o /templates

Templates are written to /tmp/templates. Copy the needed ones to this project's templates directory.

Testing

Service tests are located in gsrest/test. These are called from generated controller tests located in openapi_server/test/.

To launch the integration tests, use tox:

pip install tox
tox

This launches a mockup Cassandra database instance and ingests test data from tests/data. See tox.ini (templates/tox.mustache respectively) and tests/Makefile for further details on test configuration and setup.

Test instance configuration template can be found in tests/config.yaml.template. The test instance configuration is generated automatically from the template. You may only need to edit the template if you add more test keyspaces.

You may use tests/fetch_data.sh to fetch live data and customize for testing.

Testing in Docker

In order to have a reproducible test environment, tests can also be run via docker.

export UID  # in order to inject user id into the docker container and avoid permission issues
cd tests
make test

On OS X you need to install GNU sed and link it on your system as sed.