OpenTracing

Ariadne provides an extension that implements the OpenTracing specification, making it easy to monitor GraphQL API performance and errors using popular APM tools like Datadog or Jaeger.

Note: for performance reasons OpenTracing extension excludes default resolvers.

Enabling OpenTracing in the API

To enable OpenTracing in your API, import the OpenTracingExtension class from ariadne.contrib.tracing.opentracing and pass it to your server extensions option:

from ariadne.contrib.tracing.opentracing import OpenTracingExtension

app = GraphQL(
    schema,
    debug=True,
    extensions=[OpenTracingExtension],
)

Note: If using WSGI, use OpenTracingExtensionSync in place of OpenTracingExtension.

Note: If you don't have OpenTracing already configured in your project, you will need to install the opentracing-python package and configure tracer for your APM solution.

Sanitizing sensitive arguments data

By default all arguments are sent to the APM service. If your API fields have arguments for sensitive data like passwords or tokens, you will need to sanitize those before sending tracking data to the service.

OpenTracingExtension has single configuration option named arg_filter, that can be a function that extension will call with the copy of the dict of arguments previously passed to field's resolver.

Because extension instances are created per request, OpenTracingExtension can't be instantiated inside the extensions. Instead Ariadne provides opentracing_extension utility that creates partial function that initializes the OpenTracingExtension with arg_filter option per request.

Here is an example defining custom sanitizing function named my_arg_filter and using opentracing_extension to enable OpenTracing with it:

from ariadne.contrib.tracing import opentracing_extension

def my_arg_filter(args, info):
    if "password" in args:
        args["password"] = "[redacted]"
    if "secret" in args:
        args["secret"] = "[redacted]"
    for key, value in args.items():
        if isinstance(value, dict):
            args[key] = my_arg_filter(value)
        if isinstance(value, list):
            args[key] = [my_arg_filter(v) for v in value]
    return args


schema = make_executable_schema(type_def, [query, mutation])
app = GraphQL(
    schema,
    debug=True,
    extensions=[
        opentracing_extension(arg_filter=my_arg_filter),
    ],
)

Note: If you are using WSGI, use opentracing_extension_sync in place of opentracing_extension.