Add OpenTelemetry Sample (#61)

* copy over chad's commit

* add test and update readme

* rename module

* Update README.md

Co-authored-by: Chad Retz <chad@temporal.io>

---------

Co-authored-by: Chad Retz <chad@temporal.io>

Author: chris-olszewski

Gemfile

@@ -24,3 +24,9 @@ group :encryption do
   gem 'rackup'
   gem 'sinatra'
 end
+
+group :opentelemetry do
+  gem 'opentelemetry-api'
+  gem 'opentelemetry-exporter-otlp'
+  gem 'opentelemetry-sdk'
+end

Gemfile.lock

@@ -62,6 +62,25 @@ GEM
     oj (3.16.10)
       bigdecimal (>= 3.0)
       ostruct (>= 0.2)
+    opentelemetry-api (1.7.0)
+    opentelemetry-common (0.22.0)
+      opentelemetry-api (~> 1.0)
+    opentelemetry-exporter-otlp (0.30.0)
+      google-protobuf (>= 3.18)
+      googleapis-common-protos-types (~> 1.3)
+      opentelemetry-api (~> 1.1)
+      opentelemetry-common (~> 0.20)
+      opentelemetry-sdk (~> 1.2)
+      opentelemetry-semantic_conventions
+    opentelemetry-registry (0.4.0)
+      opentelemetry-api (~> 1.1)
+    opentelemetry-sdk (1.9.0)
+      opentelemetry-api (~> 1.1)
+      opentelemetry-common (~> 0.20)
+      opentelemetry-registry (~> 0.2)
+      opentelemetry-semantic_conventions
+    opentelemetry-semantic_conventions (1.36.0)
+      opentelemetry-api (~> 1.0)
     ostruct (0.6.1)
     parallel (1.26.3)
     parser (3.3.6.0)
@@ -134,6 +153,9 @@ PLATFORMS
 
 DEPENDENCIES
   minitest
+  opentelemetry-api
+  opentelemetry-exporter-otlp
+  opentelemetry-sdk
   puma
   rackup
   rake

README.md

@@ -27,6 +27,7 @@ Prerequisites:
 * [encryption](encryption) - Demonstrates how to make a codec for end-to-end encryption.
 * [env_config](env_config) - Load client configuration from TOML files with programmatic overrides.
 * [message_passing_simple](message_passing_simple) - Simple workflow that accepts signals, queries, and updates.
+* [open_telemetry](open_telemetry) - Demonstrates how to use OpenTelemetry tracing and metrics with the Ruby SDK
 * [patching](patching) - Demonstrates how to safely alter a workflow.
 * [polling/frequent](polling/frequent) - Implement a frequent polling mechanism inside an Activity.
 * [polling/infrequent](polling/infrequent) - Implement an infrequent polling mechanism using Temporal's automatic Activity Retry feature.

open_telemetry/README.md

@@ -0,0 +1,33 @@
+# OpenTelemetry Sample
+
+Demonstrates how to use OpenTelemetry tracing and metrics with the Ruby SDK
+
+## How to Run
+
+First, in another terminal start up a Grafana OpenTelemetry instance which will
+collect telemetry and provide the Grafana UI for viewing the data.
+
+```bash
+  docker compose up
+```
+
+In another terminal, start the worker
+
+```bash
+  bundle exec ruby worker.rb
+```
+
+Finally start the workflow
+
+```bash
+  bundle exec ruby starter.rb
+```
+
+You should be able to see the result in the terminal.
+
+To view the Grafana dashboard go to `http://localhost:3000`
+
+You can find the trace by clicking on the "Explore" tab,
+selecting "Tempo" as the data source, and switching the query type to "Search".
+
+There will be a trace for `my-service` containing the workflow trace.

open_telemetry/compose_greeting_activity.rb

@@ -0,0 +1,30 @@
+# frozen_string_literal: true
+
+require 'temporalio/activity'
+
+module OpenTelemetry
+  class ComposeGreetingActivity < Temporalio::Activity::Definition
+    def initialize(tracer)
+      @tracer = tracer
+    end
+
+    def execute(name)
+      # Capture start time for histogram metric later
+      start_time = Process.clock_gettime(Process::CLOCK_MONOTONIC)
+
+      # Run activity in our own span. Most users will not need to create their own spans in activities, they will just
+      # rely on the default spans implicitly created. This is just a sample to show it can be done.
+      @tracer.in_span('my-activity-span', attributes: { 'my-group-attr' => 'simple-activities' }) do
+        # Sleep for a second, then return
+        sleep(1)
+        "Hello, #{name}!"
+      ensure
+        # Custom metrics can be created inside activities
+        Temporalio::Activity::Context.current.metric_meter
+                                     .create_metric(:histogram, 'my-activity-histogram', value_type: :duration)
+                                     .with_additional_attributes({ 'my-group-attr' => 'simple-activities' })
+                                     .record(Process.clock_gettime(Process::CLOCK_MONOTONIC) - start_time)
+      end
+    end
+  end
+end

open_telemetry/docker-compose.yaml

@@ -0,0 +1,8 @@
+services:
+  grafana-dashboard:
+    image: grafana/otel-lgtm:latest
+    tty: true
+    ports:
+      - 3000:3000
+      - 4317:4317
+      - 4318:4318

open_telemetry/greeting_workflow.rb

@@ -0,0 +1,40 @@
+# frozen_string_literal: true
+
+require 'temporalio/workflow'
+require 'temporalio/contrib/open_telemetry'
+require_relative 'compose_greeting_activity'
+
+module OpenTelemetry
+  class GreetingWorkflow < Temporalio::Workflow::Definition
+    def initialize
+      # Custom metrics can be created inside workflows. Most users will not need to create custom metrics inside
+      # workflows, this just shows that it can be done.
+      @my_workflow_counter = Temporalio::Workflow.metric_meter.create_metric(:counter, 'my-workflow-counter')
+                                                 .with_additional_attributes({ 'my-group-attr' => 'simple-workflows' })
+    end
+
+    def execute(name)
+      # Increment our custom metric
+      @my_workflow_counter.record(35)
+
+      # We can create a span in the workflow too. This is just an example to show this can be done, most users will not
+      # create spans in workflows but rather rely on the defaults.
+      #
+      # This span is completed as soon as created because OpenTelemetry doesn't support spans that may have to be
+      # completed on different machines. The span will be parented to the outer workflow span. Whether the outer span is
+      # the "StartWorkflow" from the client or the "RunWorkflow" where it first ran depends on if this is replayed
+      # separately from where it started. See the Ruby SDK README for more details.
+      Temporalio::Contrib::OpenTelemetry::Workflow.with_completed_span(
+        'my-workflow-span',
+        attributes: { 'my-group-attr' => 'simple-workflows' }
+      ) do
+        # The span will be the parent of the span created here to start the activity
+        Temporalio::Workflow.execute_activity(
+          ComposeGreetingActivity,
+          name, # Activity argument
+          start_to_close_timeout: 5 * 60 # 5 minutes
+        )
+      end
+    end
+  end
+end

open_telemetry/starter.rb

@@ -0,0 +1,38 @@
+# frozen_string_literal: true
+
+require 'opentelemetry/sdk'
+require 'temporalio/client'
+require 'temporalio/contrib/open_telemetry'
+require 'temporalio/runtime'
+require_relative 'greeting_workflow'
+require_relative 'util'
+
+# Configure metrics and tracing
+OpenTelemetry::Util.configure_metrics_and_tracing
+
+# Demonstrate that we can create a custom metric right on the runtime, though most users won't need this
+Temporalio::Runtime.default.metric_meter.create_metric(:gauge, 'my-starter-gauge', value_type: :float)
+                   .with_additional_attributes({ 'my-group-attr' => 'simple-starters' })
+                   .record(1.23)
+
+# Create a client with the tracing interceptor set using the tracer
+tracer = OpenTelemetry.tracer_provider.tracer('temporal_ruby_sample', '0.1.0')
+client = Temporalio::Client.connect(
+  'localhost:7233',
+  'default',
+  interceptors: [Temporalio::Contrib::OpenTelemetry::TracingInterceptor.new(tracer)]
+)
+
+# Demonstrate an arbitrary outer span. Most users may not explicitly create outer spans before using clients and rather
+# solely rely on the implicit ones created in the client via interceptor, but this demonstrates that it can be done.
+tracer.in_span('my-client-span', attributes: { 'my-group-attr' => 'simple-client' }) do
+  # Run workflow
+  puts 'Executing workflow'
+  result = client.execute_workflow(
+    OpenTelemetry::GreetingWorkflow,
+    'User', # Workflow argument
+    id: 'opentelemetry-sample-workflow-id',
+    task_queue: 'opentelemetry-sample'
+  )
+  puts "Workflow result: #{result}"
+end

open_telemetry/util.rb

@@ -0,0 +1,41 @@
+# frozen_string_literal: true
+
+require 'opentelemetry/exporter/otlp'
+require 'opentelemetry/sdk'
+require 'temporalio/contrib/open_telemetry'
+require 'temporalio/runtime'
+
+module OpenTelemetry
+  module Util
+    def self.configure_metrics_and_tracing
+      # Before doing anything, configure the default runtime with OpenTelemetry metrics. Unlike OpenTelemetry tracing in
+      # Temporal, OpenTelemetry metrics does not use the Ruby OpenTelemetry library, but rather an internal one.
+      Temporalio::Runtime.default = Temporalio::Runtime.new(
+        telemetry: Temporalio::Runtime::TelemetryOptions.new(
+          metrics: Temporalio::Runtime::MetricsOptions.new(
+            opentelemetry: Temporalio::Runtime::OpenTelemetryMetricsOptions.new(
+              url: 'http://127.0.0.1:4317',
+              durations_as_seconds: true
+            )
+          )
+        )
+      )
+      # Globally configure the Ruby OpenTelemetry library for tracing purposes. As of this writing, OpenTelemetry Ruby
+      # does not support OTLP over gRPC, so we use the HTTP endpoint instead.
+      OpenTelemetry::SDK.configure do |c|
+        c.service_name = 'my-service'
+        c.use_all
+        # Can use a SimpleSpanProcessor instead of a BatchSpanProcessor, but batch is better for production and moves
+        # the span exporting outside of the workflow instead of synchronously inside the workflow context.
+        processor = OpenTelemetry::SDK::Trace::Export::BatchSpanProcessor.new(
+          OpenTelemetry::Exporter::OTLP::Exporter.new(
+            endpoint: 'http://localhost:4318/v1/traces'
+          )
+        )
+        c.add_span_processor(processor)
+        # We need to shutdown the batch span processor on process exit to flush spans
+        at_exit { processor.shutdown }
+      end
+    end
+  end
+end

open_telemetry/worker.rb

@@ -0,0 +1,36 @@
+# frozen_string_literal: true
+
+require 'opentelemetry/sdk'
+require 'temporalio/client'
+require 'temporalio/contrib/open_telemetry'
+require 'temporalio/runtime'
+require 'temporalio/worker'
+require_relative 'compose_greeting_activity'
+require_relative 'greeting_workflow'
+require_relative 'util'
+
+# Configure metrics and tracing
+OpenTelemetry::Util.configure_metrics_and_tracing
+
+# Demonstrate that we can create a custom metric right on the runtime, though most users won't need this
+Temporalio::Runtime.default.metric_meter.create_metric(:gauge, 'my-worker-gauge', value_type: :float)
+                   .with_additional_attributes({ 'my-group-attr' => 'simple-workers' })
+                   .record(1.23)
+
+# Create a client with the tracing interceptor set using the tracer
+tracer = OpenTelemetry.tracer_provider.tracer('opentelemetry_sample', '1.0.0')
+client = Temporalio::Client.connect(
+  'localhost:7233',
+  'default',
+  interceptors: [Temporalio::Contrib::OpenTelemetry::TracingInterceptor.new(tracer)]
+)
+
+# Run worker
+worker = Temporalio::Worker.new(
+  client:,
+  task_queue: 'opentelemetry-sample',
+  activities: [OpenTelemetry::ComposeGreetingActivity.new(tracer)],
+  workflows: [OpenTelemetry::GreetingWorkflow]
+)
+puts 'Starting worker (ctrl+c to exit)'
+worker.run(shutdown_signals: ['SIGINT'])