This module provides automatic instrumentation and tracing for GraphQL in Node.js applications, which may be loaded using the @opentelemetry/sdk-trace-node package and is included in the @opentelemetry/auto-instrumentations-node bundle.
If total installation size is not constrained, it is recommended to use the @opentelemetry/auto-instrumentations-node bundle with @opentelemetry/sdk-node for the most seamless instrumentation experience.
Note: graphql plugin instruments graphql directly. it should work with any package that wraps the graphql package (e.g apollo).
Compatible with OpenTelemetry JS API and SDK 1.0+.
npm install @opentelemetry/instrumentation-graphql^14.0 | ^15.0 | ^16.0
'use strict';
const { GraphQLInstrumentation } = require('@opentelemetry/instrumentation-graphql');
const { NodeTracerProvider } = require('@opentelemetry/sdk-trace-node');
const { registerInstrumentations } = require('@opentelemetry/instrumentation');
const provider = new NodeTracerProvider();
provider.register();
registerInstrumentations({
instrumentations: [
new GraphQLInstrumentation({
// optional params
// allowValues: true,
// depth: 2,
// mergeItems: true,
// ignoreTrivialResolveSpans: true,
// ignoreResolveSpans: true,
}),
],
});| Param | type | Default Value | Description | |
|---|---|---|---|---|
| mergeItems | boolean | false | Whether to merge list items into a single element. example: users.*.name instead of users.0.name, users.1.name
|
|
| depth | number | -1 | The maximum depth of fields/resolvers to instrument. When set to 0 it will not instrument fields and resolvers. When set to -1 it will instrument all fields and resolvers. | |
| allowValues | boolean | false | When set to true it will not remove attributes values from schema source. By default all values that can be sensitive are removed and replaced with "*" | |
| ignoreTrivialResolveSpans | boolean | false | Don't create spans for the execution of the default resolver on object properties. | |
| ignoreResolveSpans | boolean | false | Don't create spans for resolvers, regardless if they are trivial or not. | |
| responseHook | GraphQLInstrumentationExecutionResponseHook | undefined | Hook that allows adding custom span attributes based on the data returned from "execute" GraphQL action. |
The instrumentation by default will create a span for each invocation of a resolver.
A resolver is run by graphql for each field in the query response, which can be a lot of spans for objects with many properties, or when lists are involved.
There are few config options which can be used to reduce the verbosity of the instrumentations.
They are all disabled by default. User can opt in to any combination of them to control the amount of spans.
When a resolver function is not defined on the schema for a field, graphql will use the default resolver which just looks for a property with that name on the object. If the property is not a function, it's not very interesting to trace.
The performance overhead for complex schemas with a lot of resolvers can be high due to the large number of spans created. When ignoreResolveSpans is set to true, no spans for resolvers will be created.
If you are using @apollo/server as your graphql server, you might want to
enable this option because all resolvers are currently considered non-trivial.
The depth is the number of nesting levels of the field, and the following is a query with a depth of 3:
{
a {
b {
c
}
}
}You can limit the instrumentation to stop recording "resolve" spans after a specific depth is reached.
-
-1means no limit. -
0means don't record any "resolve" spans. -
2for the example above will record a span for resolving "a" and "b" but not "c".
When resolving a field to a list, graphql will execute a resolver for every field in the query on every object in the list.
When setting mergeItems to true it will only record a span for the first invocation of a resolver on each field in the list, marking it's path as "foo.*.bar" instead of "foo.0.bar", "foo.1.bar".
Notice that all span data only reflects the invocation on the first element. That includes timing, events and status.
Downstream spans in the context of all resolvers will be child of the first span.
Can be found here
- For more information on OpenTelemetry, visit: https://opentelemetry.io/
- For more about OpenTelemetry JavaScript: https://github.com/open-telemetry/opentelemetry-js
- For help or feedback on this project, join us in GitHub Discussions
Apache 2.0 - See LICENSE for more information.