Setup:

Create a virtual env using venv, for example:

python3 -m venv .dgraph-locust && \ 
source .dgraph-locust/bin/activate && \ 
cd dgraph-locust  

Install dependencies from requirements.txt using pip :

pip install -r requirements.txt

Setup a Dgraph cluster quickly using docker compose:

docker compose up -d 

Upload the schema and load test data:

python3 common/seed.py

This test data (available in the data dir) was generated using the Dgraph datagen tool which can be downloaded here. Please refer the schema.graphql for more info on that data-model.

Run Tests:

Add queries and mutations to the queries.txt and mutations.txt files respectively. The files in this repo contain sample queries corresponding to the test-data, but all queries/mutations must be added to these files.

NOTE: Please ensure that each query/mutation is separated by a newline which the parser relies on as a delimiter, to identify separate queries or mutations. The parsing will fail if the newlines are not added or are included within queries themselves.

With Web-UI:

To start the query load-tests:

locust -f locustfiles/query.py

The Locust Web UI can be accessed at http://localhost:8089.

To start mutation load-tests:

locust -f locustfiles/mutation.py

To start a mixed-load test comprising of both queries and mutations:

locust -f locustfiles/mixed_load.py

Without WebUI:

locust -f locustfiles/query.py --headless -H http://localhost:8080 -u 1000 -r 100 --run-time 3h30m

Project Structure:

common

• The seed.py is used to upload the GraphQL schema and load data using the Dgraph Live Loader.

• The job.py is used to setup and collect Go profiles for CPU, heap and goroutines. The files are saved in the output dir, within the respective sub-directories.

• The setup.py file contains logic to retrieve data to be used as filter values in the queries to test.

• The helpers.py contains common reusable methods for the tests.

locustfiles

• Contains the locust test files that can be run as noted above.

output

• Contains sub-directories for storing profiling info.