Tracing

Every production API needs to be monitored and traced. Strawberry provides integration for Datadog and OpenTracing. Tracing is disabled by default, but can be enabled by adding an extension when creating your schema. In api/schema.py you can add the Datadog tracing extension like this:

from strawberry.extensions.tracing import DatadogTracingExtension

schema = strawberry.Schema(
    query=Query,
    mutation=Mutation,
    extensions=[
        DatadogTracingExtension,
    ],
)

Strawberry extension

Strawberry has a concept of extensions that can be used to add custom logic to the execution of a query. Extensions are useful to add tracing, logging, or any other custom logic.

The Datadog tracing extension hooks into the request start, request end and resolve functions. It will allow to trace the execution of a query and send the data to Datadog.

Example

Getting up and running with datadog or other tracing tools takes a bit of time, so I'll share a screenshot of what you'll see in the dashboard:

This view will allow to see the execution time of each field, and the time spend executing the whole query. You'll also be able to setup filters and metrics on the dashboard to get more insights on your API.

Operation names in GraphQL

We haven't talked about operation names yet, but they are an important part of GraphQL. Operation names are used to identify the query or mutation that is being executed. They are useful to group queries together, and to identify queries that are slow using tools like datadog.

For example if we use a query like this:

query {
  latestEpisodes {
    title
    podcast {
      title
    }
  }
}

we aren't really able to identify the query in datadog, or to distinguish it from similar queries.

To solve this problem we can add an operation name to the query:

query LatestEpisodes {
  latestEpisodes {
    title
    podcast {
      title
    }
  }
}

In this case we named our query LatestEpisodes, and we can now use this name to identify the query in datadog.