GraphQL Metrics

Extract as much much detail as you want from GraphQL queries, served up from your Ruby app and the graphql gem. Compatible with the graphql-batch gem, to extract batch-loaded fields resolution timings.

Be sure to read the CHANGELOG to stay updated on feature additions, breaking changes made to this gem.

NOTE: Not tested with graphql-ruby's multiplexing feature. Metrics may not be accurate if you execute multiple operations at once.

Installation

Add this line to your application's Gemfile:

gem 'graphql-metrics'

You can require it with in your code as needed with:

require 'graphql/metrics'

Or globally in the Gemfile with:

gem 'graphql-metrics', require: 'graphql/metrics'

And then execute:

$ bundle

Or install it yourself as:

$ gem install graphql-metrics

Usage

Get started by defining your own Analyzer, inheriting from GraphQL::Metrics::Analyzer.

The following analyzer demonstrates a simple way to capture commonly used metrics sourced from key parts of your schema definition, the query document being served, as well as runtime query and resolver timings. In this toy example, all of this data is simply stored on the GraphQL::Query context, under a namespace to avoid collisions with other analyzers etc.

What you do with these captured metrics is up to you!

NOTE: If any non-graphql-ruby gem-related exceptions occur in your application during query document parsing and validation, runtime metrics for queries (like query_duration, parsing_start_time_offset etc.) as well as field resolver timings (like resolver_timings, lazy_resolver_timings) may not be present in the extracted metrics hash, even if you opt to collect them by using GraphQL::Metrics::Analyzer and GraphQL::Metrics::Tracer.

Define your own analyzer subclass

  class SimpleAnalyzer < GraphQL::Metrics::Analyzer
    ANALYZER_NAMESPACE = :simple_analyzer_namespace

    def initialize(query_or_multiplex)
      super

      # `query` is defined on instances of objects inheriting from GraphQL::Metrics::Analyzer
      ns = query.context.namespace(ANALYZER_NAMESPACE)
      ns[:simple_extractor_results] = {}
    end

    # @param metrics [Hash] Query metrics, including a few details about the query document itself, as well as runtime
    # timings metrics, intended to be compatible with the Apollo Tracing spec:
    # https://github.com/apollographql/apollo-tracing#response-format
    #
    # {
    #   operation_type: "query",
    #   operation_name: "PostDetails",
    #   query_start_time: 1573833076.027327,
    #   query_duration: 2.0207119999686256,
    #   lexing_start_time_offset: 0.0010339999571442604,
    #   lexing_duration: 0.0008190000080503523,
    #   parsing_start_time_offset: 0.0010339999571442604,
    #   parsing_duration: 0.0008190000080503523,
    #   validation_start_time_offset: 0.0030819999519735575,
    #   validation_duration: 0.01704599999357015,
    #   analysis_start_time_offset: 0.0010339999571442604,
    #   analysis_duration: 0.0008190000080503523,
    #   multiplex_start_time: 0.0008190000080503523,
    # }
    #
    # You can use these metrics to track high-level query performance, along with any other details you wish to
    # manually capture from `query` and/or `query.context`.
    def query_extracted(metrics)
      custom_metrics_from_context = {
        request_id: query.context[:request_id],
        # ...
      }

      # You can make use of captured metrics here (logging to Kafka, request logging etc.)
      # log_metrics(:fields, metrics)
      #
      # Or store them on the query context:
      store_metrics(:queries, metrics.merge(custom_metrics_from_context))
    end

    # For use after controller:
    # class GraphQLController < ActionController::Base
    #   def graphql_query
    #     query_result = graphql_query.result.to_h
    #     do_something_with_metrics(query.context[:simple_extractor_results])
    #     render json: graphql_query.result
    #   end
    # end

    # @param metrics [Hash] Field selection metrics, including resolver timings metrics, also adhering to the Apollo
    # Tracing spec referred to above.
    #
    # `resolver_timings` is populated any time a field is resolved (which may be many times, if the field is nested
    # within a list field e.g. a Relay connection field).
    #
    # `lazy_resolver_timings` is only populated by fields that are resolved lazily (for example using the
    # graphql-batch gem) or that are otherwise resolved with a Promise. Any time spent in the field's resolver to
    # prepare work to be done "later" in a Promise, or batch loader will be captured in `resolver_timings`. The time
    # spent actually doing lazy field loading, including time spent within a batch loader can be obtained from
    # `lazy_resolver_timings`.
    #
    # {
    #   field_name: "id",
    #   return_type_name: "ID",
    #   parent_type_name: "Post",
    #   deprecated: false,
    #   path: ["post", "id"],
    #   resolver_timings: [
    #     start_time_offset: 0.011901999998372048,
    #     duration: 5.999987479299307e-06
    #   ],
    #   lazy_resolver_timings: [
    #     start_time_offset: 0.031901999998372048,
    #     duration: 5.999987479299307e-06
    #   ],
    # }
    def field_extracted(metrics)
      store_metrics(:fields, metrics)
    end

    # @param metrics [Hash] Argument usage metrics, including a few details about the query document itself, as well
    # as resolver timings metrics, also ahering to the Apollo Tracing spec referred to above.
    # {
    #   argument_name: "id",
    #   argument_type_name: "ID",
    #   parent_field_name: "post",
    #   parent_field_type_name: "QueryRoot",
    #   default_used: false,
    #   value_is_null: false,
    #   value: <GraphQL::Query::Arguments::ArgumentValue>,
    # }
    #
    # `value` is exposed here, in case you want to get access to the argument's definition, including the type
    # class which defines it, e.g. `metrics[:value].definition.metadata[:type_class]`
    def argument_extracted(metrics)
      store_metrics(:arguments, metrics)
    end

    private

    def store_metrics(context_key, metrics)
      ns = query.context.namespace(ANALYZER_NAMESPACE)
      ns[:simple_extractor_results][context_key] ||= []
      ns[:simple_extractor_results][context_key] << metrics
    end
  end

Once defined, you can opt into capturing all metrics seen above by simply including GraphQL::Metrics as a plugin on your schema.

Make use of your analyzer

Ensure that your schema is using the graphql-ruby 1.9+ GraphQL::Execution::Interpreter and GraphQL::Analysis::AST engine, and then simply add the below GraphQL::Metrics plugins.

This opts you in to capturing all static and runtime metrics seen above.

class Schema < GraphQL::Schema
  query QueryRoot
  mutation MutationRoot

  query_analyzer SimpleAnalyzer

  instrument :query, GraphQL::Metrics::Instrumentation.new # Both of these are required if either is used.
  tracer GraphQL::Metrics::Tracer.new                      # <-- Note!

  use GraphQL::Batch # Optional, but highly recommended. See https://github.com/Shopify/graphql-batch/.
end

Optionally, only gather static metrics

If you don't care to capture runtime metrics like query and resolver timings, you can use your analyzer a standalone analyzer without GraphQL::Metrics::Instrumentation and tracer GraphQL::Metrics::Tracer, like so:

class Schema < GraphQL::Schema
  query QueryRoot
  mutation MutationRoot

  query_analyzer SimpleAnalyzer
end

Your analyzer will still be called with query_extracted, field_extracted, but with timings metrics omitted. argument_extracted will work exactly the same, whether instrumentation and tracing are used or not.

Order of execution

Because of the structure of graphql-ruby's plugin architecture, it may be difficult to build an intuition around the order in which methods defined on GraphQL::Metrics::Instrumentation, GraphQL::Metrics::Tracer and subclasses of GraphQL::Metrics::Analyzer run.

Although you ideally will not need to care about these details if you are simply using this gem to gather metrics in your application as intended, here's a breakdown of the order of execution of the methods involved:

When used as instrumentation, an analyzer and tracing, the order of execution is:

• Tracer.capture_lexing_time
• Tracer.capture_parsing_time
• Tracer.capture_multiplex_start_time
• Instrumentation.before_query (context setup)
• Tracer.capture_validation_time
• Tracer.capture_analysis_time
• Analyzer#initialize (bit more context setup, instance vars setup)
• Analyzer#result
• Tracer.capture_query_start_time
• Tracer.trace_field (n times)
• Instrumentation.after_query (call query and field callbacks, now that we have all static and runtime metrics gathered)
• Analyzer#extract_query
• Analyzer#query_extracted
• Analyzer#extract_fields_with_runtime_metrics
  • calls Analyzer#field_extracted n times

When used as a simple analyzer, which doesn't gather or emit any runtime metrics (timings, arg values):

• Analyzer#initialize
• Analyzer#field_extracted n times
• Analyzer#result
• Analyzer#extract_query
• Analyzer#query_extracted

Development

After checking out the repo, run bin/setup to install dependencies. Then, run bundle exec rake test to run the tests. You can also run bin/console for an interactive prompt that will allow you to experiment.

To install this gem onto your local machine, run bundle exec rake install. To release a new version, update the version number in version.rb, and then run bundle exec rake release, which will create a git tag for the version, push git commits and tags, and push the .gem file to rubygems.org.

Contributing

Bug reports and pull requests are welcome on GitHub at https://github.com/Shopify/graphql-metrics. This project is intended to be a safe, welcoming space for collaboration, and contributors are expected to adhere to the Contributor Covenant code of conduct.

License

The gem is available as open source under the terms of the MIT License.

Code of Conduct

Everyone interacting in the GraphQL::Metrics project’s codebases, issue trackers, chat rooms and mailing lists is expected to follow the code of conduct.