The Query Service interfaces with Apache Pinot Data Store

Hypertrace Query Architecture

• Query Service serves time series data for attributes/metrics from spans and events. The query interface this exposes is more like a table where you can select any columns, aggregations on them with filters. It's easy to do slicing and dicing of data (only from one table since no JOINs are supported) with this interface.
• Currently Pinot is the only DB layer for this service and queries are directly translated to PQL. However, in future we could add support for more data stores like Presto, Druid etc here.
• This layer doesn't understand the entities and their relationships because the query interface is generic table like interface.

The Query service uses gradlew to compile/install/distribute. Gradle wrapper is already part of the source code. To build Query Service, run:

./gradlew dockerBuildImages

Run ./gradlew test to execute unit tests.

To test your image using the docker-compose setup follow the steps:

• Commit you changes to a branch say query-service-test.
• Go to hypertrace-service and checkout the above branch in the submodule.

cd query-service && git checkout query-service-test && cd ..

• Change tag for hypertrace-service from :main to :test in docker-compose file like this.

  hypertrace-service:
    image: hypertrace/hypertrace-service:test
    container_name: hypertrace-service
    ...

• and then run docker-compose up to test the setup.

and then run docker-compose up to test the setup.

Add image repository and tag in values.yaml file here like below and then run ./hypertrace.sh install again and you can test your image!

query-service:
  image:
    repository: "hypertrace/query-service"
    tagOverride: "test"

• DockerHub > Query service