ddtrace 0.48.0 → 0.51.1

data/benchmarks/profiler_sample_loop.rb

@@ -0,0 +1,69 @@
+require 'benchmark/ips'
+require 'ddtrace'
+require 'pry'
+
+# This benchmark measures the performance of the main stack sampling loop of the profiler
+
+class ProfilerSampleLoopBenchmark
+  def create_profiler
+    Datadog.configure do |c|
+      # c.diagnostics.debug = true
+      c.profiling.enabled = true
+      c.tracer.transport_options = proc { |t| t.adapter :test }
+    end
+
+    # Stop background threads
+    Datadog.profiler.shutdown!
+
+    # Call collection directly
+    @stack_collector = Datadog.profiler.collectors.first
+    @recorder = @stack_collector.recorder
+  end
+
+  def thread_with_very_deep_stack(depth: 500)
+    deep_stack = proc do |n|
+      if n > 0
+        deep_stack.call(n - 1)
+      else
+        sleep
+      end
+    end
+
+    Thread.new { deep_stack.call(depth) }.tap { |t| t.name = "Deep stack #{depth}" }
+  end
+
+  def run_benchmark
+    Benchmark.ips do |x|
+      x.config(time: 10, warmup: 2)
+
+      x.report("stack collector #{ENV['CONFIG']}") do
+        @stack_collector.collect_and_wait
+      end
+
+      x.save! 'profiler-sample-loop-results'
+      x.compare!
+    end
+
+    @recorder.flush
+  end
+
+  def run_forever
+    while true
+      1000.times { @stack_collector.collect_and_wait }
+      @recorder.flush
+      print '.'
+    end
+  end
+end
+
+puts "Current pid is #{Process.pid}"
+
+ProfilerSampleLoopBenchmark.new.instance_exec do
+  create_profiler
+  4.times { thread_with_very_deep_stack }
+  if ARGV.include?('--forever')
+    run_forever
+  else
+    run_benchmark
+  end
+end

data/benchmarks/profiler_submission.rb

@@ -0,0 +1,93 @@
+require 'benchmark/ips'
+require 'ddtrace'
+require 'pry'
+
+# This benchmark measures the performance of encoding pprofs and trying to submit them
+#
+# The FLUSH_DUMP_FILE (by default benchmarks/data/profiler-submission-marshal.gz, gathered from benchmarking using
+# the discourse forum rails app) can be generated by changing the scheduler.rb#flush_events to dump the contents of
+# "flush" to a file during a benchmark execution:
+#
+# dump_file = "marshal-#{Time.now.utc.to_i}.dump"
+# File.open(dump_file, "w") { |f| Marshal.dump(flush, f) }
+# Datadog.logger.info("Dumped to #{dump_file}")
+#
+# And then gzipping the result. (This can probably be automated a bit by adding an extra exporter, but the above worked
+# for me).
+
+class ProfilerSubmission
+  def create_profiler
+    @adapter_buffer = []
+
+    Datadog.configure do |c|
+      # c.diagnostics.debug = true
+      c.profiling.enabled = true
+      c.tracer.transport_options = proc { |t| t.adapter :test, @adapter_buffer }
+    end
+
+    # Stop background threads
+    Datadog.profiler.shutdown!
+
+    # Call exporter directly
+    @exporter = Datadog.profiler.scheduler.exporters.first
+    @flush = Marshal.load(
+      Zlib::GzipReader.new(File.open(ENV['FLUSH_DUMP_FILE'] || 'benchmarks/data/profiler-submission-marshal.gz'))
+    )
+  end
+
+  def check_valid_pprof
+    output_pprof = @adapter_buffer.last[:form]["data[0]"].io
+
+    expected_hashes = [
+      "75c65dec2d5d750faca3e905486c39ba44219636ac1bea15d5d624b741d4b62a",
+      "6780c47b3e271f2abe0346fd40ac358f6a34270b3a8f5743aacd1970bbbbc6f5"
+    ]
+    current_hash = Digest::SHA256.hexdigest(Zlib::GzipReader.new(output_pprof).read)
+
+    if expected_hashes.include?(current_hash)
+      puts "Output hash #{current_hash} matches known signature"
+    else
+      puts "WARNING: Unexpected pprof output -- unknown hash. Hashes seem to differ due to some of our dependencies changing, " \
+        "but it can also indicate that encoding output has become corrupted."
+    end
+  end
+
+  def run_benchmark
+    Benchmark.ips do |x|
+      x.config(time: 10, warmup: 2)
+
+      x.report("exporter #{ENV['CONFIG']}") do
+        run_once
+      end
+
+      x.save! 'profiler-submission-results.json'
+      x.compare!
+    end
+  end
+
+  def run_forever
+    while true
+      run_once
+      print '.'
+    end
+  end
+
+  def run_once
+    @adapter_buffer.clear
+    @exporter.export(@flush)
+  end
+end
+
+puts "Current pid is #{Process.pid}"
+
+ProfilerSubmission.new.instance_exec do
+  create_profiler
+  run_once
+  check_valid_pprof
+
+  if ARGV.include?('--forever')
+    run_forever
+  else
+    run_benchmark
+  end
+end

data/ddtrace.gemspec

@@ -7,7 +7,7 @@ require 'ddtrace/version'
 Gem::Specification.new do |spec|
   spec.name                  = 'ddtrace'
   spec.version               = Datadog::VERSION::STRING
-  spec.required_ruby_version = ">= #{Datadog::VERSION::MINIMUM_RUBY_VERSION}"
+  spec.required_ruby_version = [">= #{Datadog::VERSION::MINIMUM_RUBY_VERSION}", "< #{Datadog::VERSION::MAXIMUM_RUBY_VERSION}"]
   spec.required_rubygems_version = '>= 2.0.0'
   spec.authors               = ['Datadog, Inc.']
   spec.email                 = ['dev@datadoghq.com']
@@ -35,7 +35,7 @@ Gem::Specification.new do |spec|
   if RUBY_VERSION >= '2.2.0'
     spec.add_dependency 'msgpack'
   else
-    # msgpack 1.4 fails for Ruby 2.0 and 2.1: https://github.com/msgpack/msgpack-ruby/issues/205
+    # msgpack 1.4 fails for Ruby 2.1: https://github.com/msgpack/msgpack-ruby/issues/205
     spec.add_dependency 'msgpack', '< 1.4'
   end
 

data/docker-compose.yml

@@ -1,38 +1,8 @@
 version: '3.2'
 services:
   # MRI
-  tracer-2.0:
-    image: palazzem/docker-library:ddtrace_rb_2_0_0
-    command: /bin/bash
-    depends_on:
-      - ddagent
-      - elasticsearch
-      - memcached
-      - mongodb
-      - mysql
-      - postgres
-      - presto
-      - redis
-    env_file: ./.env
-    environment:
-      - BUNDLE_GEMFILE=/app/Gemfile
-      - DD_AGENT_HOST=ddagent
-      - TEST_DATADOG_INTEGRATION=1
-      - TEST_ELASTICSEARCH_HOST=elasticsearch
-      - TEST_MEMCACHED_HOST=memcached
-      - TEST_MONGODB_HOST=mongodb
-      - TEST_MYSQL_HOST=mysql
-      - TEST_POSTGRES_HOST=postgres
-      - TEST_PRESTO_HOST=presto
-      - TEST_REDIS_HOST=redis
-    stdin_open: true
-    tty: true
-    volumes:
-      - .:/app
-      - bundle-2.0:/usr/local/bundle
-      - gemfiles-2.0:/app/gemfiles
   tracer-2.1:
-    image: palazzem/docker-library:ddtrace_rb_2_1_10
+    image: ivoanjo/docker-library:ddtrace_rb_2_1_10
     command: /bin/bash
     depends_on:
       - ddagent
@@ -61,9 +31,8 @@ services:
     volumes:
       - .:/app
       - bundle-2.1:/usr/local/bundle
-      - gemfiles-2.1:/app/gemfiles
   tracer-2.2:
-    image: palazzem/docker-library:ddtrace_rb_2_2_10
+    image: ivoanjo/docker-library:ddtrace_rb_2_2_10
     command: /bin/bash
     depends_on:
       - ddagent
@@ -92,9 +61,8 @@ services:
     volumes:
       - .:/app
       - bundle-2.2:/usr/local/bundle
-      - gemfiles-2.2:/app/gemfiles
   tracer-2.3:
-    image: palazzem/docker-library:ddtrace_rb_2_3_8
+    image: ivoanjo/docker-library:ddtrace_rb_2_3_8
     command: /bin/bash
     depends_on:
       - ddagent
@@ -123,9 +91,8 @@ services:
     volumes:
       - .:/app
       - bundle-2.3:/usr/local/bundle
-      - gemfiles-2.3:/app/gemfiles
   tracer-2.4:
-    image: palazzem/docker-library:ddtrace_rb_2_4_6
+    image: ivoanjo/docker-library:ddtrace_rb_2_4_10
     command: /bin/bash
     depends_on:
       - ddagent
@@ -154,9 +121,8 @@ services:
     volumes:
       - .:/app
       - bundle-2.4:/usr/local/bundle
-      - gemfiles-2.4:/app/gemfiles
   tracer-2.5:
-    image: marcotc/docker-library:ddtrace_rb_2_5_6
+    image: ivoanjo/docker-library:ddtrace_rb_2_5_9
     command: /bin/bash
     depends_on:
       - ddagent
@@ -185,9 +151,8 @@ services:
     volumes:
       - .:/app
       - bundle-2.5:/usr/local/bundle
-      - gemfiles-2.5:/app/gemfiles
   tracer-2.6:
-    image: marcotc/docker-library:ddtrace_rb_2_6_4
+    image: ivoanjo/docker-library:ddtrace_rb_2_6_7
     command: /bin/bash
     depends_on:
       - ddagent
@@ -216,9 +181,8 @@ services:
     volumes:
       - .:/app
       - bundle-2.6:/usr/local/bundle
-      - gemfiles-2.6:/app/gemfiles
   tracer-2.7:
-    image: marcotc/docker-library:ddtrace_rb_2_7_0
+    image: ivoanjo/docker-library:ddtrace_rb_2_7_3
     command: /bin/bash
     depends_on:
       - ddagent
@@ -247,9 +211,8 @@ services:
     volumes:
       - .:/app
       - bundle-2.7:/usr/local/bundle
-      - gemfiles-2.7:/app/gemfiles
   tracer-3.0:
-    image: marcotc/docker-library:ddtrace_rb_3.0.0
+    image: ivoanjo/docker-library:ddtrace_rb_3_0_1
     command: /bin/bash
     depends_on:
       - ddagent
@@ -275,10 +238,9 @@ services:
     volumes:
       - .:/app
       - bundle-3.0:/usr/local/bundle
-      - gemfiles-3.0:/app/gemfiles
   # JRuby
   tracer-jruby-9.2.0.0:
-    image: marcotc/docker-library:ddtrace_rb_jruby_9_2_0_0
+    image: ivoanjo/docker-library:ddtrace_rb_jruby_9_2_0_0
     command: /bin/bash
     depends_on:
       - ddagent
@@ -307,9 +269,8 @@ services:
     volumes:
       - .:/app
       - bundle-jruby-9.2.0.0:/usr/local/bundle
-      - gemfiles-jruby-9.2.0.0:/app/gemfiles
   tracer-jruby-9.2-latest:
-    image: marcotc/docker-library:ddtrace_rb_jruby_9_2_11_1
+    image: ivoanjo/docker-library:ddtrace_rb_jruby_9_2_18_0
     command: /bin/bash
     depends_on:
       - ddagent
@@ -338,10 +299,9 @@ services:
     volumes:
       - .:/app
       - bundle-jruby-9.2-latest:/usr/local/bundle
-      - gemfiles-jruby-9.2-latest:/app/gemfiles
   # TruffleRuby
-  tracer-truffleruby-21.0.0:
-    image: ivoanjo/docker-library:ddtrace_rb_truffleruby_21_0_0
+  tracer-truffleruby-21.1.0:
+    image: ivoanjo/docker-library:ddtrace_rb_truffleruby_21_1_0
     command: /bin/bash
     depends_on:
       - ddagent
@@ -369,8 +329,7 @@ services:
     tty: true
     volumes:
       - .:/app
-      - bundle-truffleruby-21.0.0:/usr/local/bundle
-      - gemfiles-truffleruby-21.0.0:/app/gemfiles
+      - bundle-truffleruby-21.1.0:/usr/local/bundle
   ddagent:
     image: datadog/agent
     environment:
@@ -442,7 +401,6 @@ services:
     ports:
       - "${TEST_REDIS_PORT}:6379"
 volumes:
-  bundle-2.0:
   bundle-2.1:
   bundle-2.2:
   bundle-2.3:
@@ -453,16 +411,4 @@ volumes:
   bundle-3.0:
   bundle-jruby-9.2.0.0:
   bundle-jruby-9.2-latest:
-  bundle-truffleruby-21.0.0:
-  gemfiles-2.0:
-  gemfiles-2.1:
-  gemfiles-2.2:
-  gemfiles-2.3:
-  gemfiles-2.4:
-  gemfiles-2.5:
-  gemfiles-2.6:
-  gemfiles-2.7:
-  gemfiles-3.0:
-  gemfiles-jruby-9.2.0.0:
-  gemfiles-jruby-9.2-latest:
-  gemfiles-truffleruby-21.0.0:
+  bundle-truffleruby-21.1.0:

data/docs/GettingStarted.md

@@ -85,6 +85,10 @@ To contribute, check out the [contribution guidelines][contribution docs] and [d
      - [Metrics](#metrics)
          - [For application runtime](#for-application-runtime)
      - [OpenTracing](#opentracing)
+     - [Profiling](#profiling)
+ - [Known issues and suggested configurations](#known-issues-and-suggested-configurations)
+    - [Payload too large](#payload-too-large)
+    - [Stack level too deep](#stack-level-too-deep)
 
 ## Compatibility
 
@@ -100,7 +104,7 @@ To contribute, check out the [contribution guidelines][contribution docs] and [d
 |       |                            | 2.3     | Full                                 | Latest              |
 |       |                            | 2.2     | Full                                 | Latest              |
 |       |                            | 2.1     | Full                                 | Latest              |
-|       |                            | 2.0     | Deprecated                           | < 0.50.0            |
+|       |                            | 2.0     | EOL since June 7th, 2021             | < 0.50.0            |
 |       |                            | 1.9.3   | EOL since August 6th, 2020           | < 0.27.0            |
 |       |                            | 1.9.1   | EOL since August 6th, 2020           | < 0.27.0            |
 | JRuby | https://www.jruby.org      | 9.2     | Full                                 | Latest              |
@@ -546,7 +550,7 @@ Datadog.configure do |c|
   # Symbol matching your database connection in config/database.yml
   # Only available if you are using Rails with ActiveRecord.
   c.use :active_record, describes: :secondary_database, service_name: 'secondary-db'
-  
+
   # Block configuration pattern.
   c.use :active_record, describes: :secondary_database do |second_db|
     second_db.service_name = 'secondary-db'
@@ -584,7 +588,7 @@ Datadog.configure do |c|
 
   # Matches any `mysql2` connection.
   c.use :active_record, describes: { adapter: 'mysql2'}, service_name: 'mysql-db'
-  
+
   # Matches any `mysql2` connection to the `reports` database.
   #
   # In case of multiple matching `describe` configurations, the latest one applies.
@@ -1565,7 +1569,7 @@ Datadog.configure do |c|
   # For network connections, only these fields are considered during matching:
   # scheme, host, port, db
   # Other fields are ignored.
-  
+
   # Network connection string
   c.use :redis, describes: 'redis://127.0.0.1:6379/0', service_name: 'redis-connection-string'
   c.use :redis, describes: { url: 'redis://127.0.0.1:6379/1' }, service_name: 'redis-connection-url'
@@ -1678,8 +1682,6 @@ Where `options` is an optional `Hash` that accepts the following parameters:
 | --- | ----------- | ------- |
 | `service_name` | Service name for `sequel` instrumentation | Name of database adapter (e.g. `'mysql2'`) |
 
-Only Ruby 2.0+ is supported.
-
 **Configuring databases to use different settings**
 
 If you use multiple databases with Sequel, you can give each of them different settings by configuring their respective `Sequel::Database` objects:
@@ -1867,7 +1869,7 @@ Datadog.configure do |c|
 
   # Breaks down very large traces into smaller batches
   c.tracer.partial_flush.enabled = false
-  
+
   # You can specify your own tracer
   c.tracer.instance = Datadog::Tracer.new
 
@@ -1954,7 +1956,7 @@ We recommend setting the environment variable `DD_TRACE_SAMPLE_RATE=1.0` in all
 App Analytics, previously configured with the `analytics_enabled` setting, is deprecated in favor of Tracing without Limits™. Documentation for this [deprecated configuration is still available](https://docs.datadoghq.com/tracing/legacy_app_analytics/).
 
 #### Application-side sampling
- 
+
 While the trace agent can sample traces to reduce bandwidth usage, application-side sampling reduces the performance overhead.
 
 This will **reduce visibility and is not recommended**. See [DD_TRACE_SAMPLE_RATE](#environment-variables) for the recommended sampling approach.
@@ -2115,7 +2117,7 @@ For more details on how to activate distributed tracing for integrations, see th
 
 - [Excon](#excon)
 - [Faraday](#faraday)
-- [Rest Client](#restclient)
+- [Rest Client](#rest-client)
 - [Net/HTTP](#nethttp)
 - [Rack](#rack)
 - [Rails](#rails)
@@ -2238,7 +2240,7 @@ In many cases, such as logging, it may be useful to correlate trace IDs to other
 
 ##### Automatic
 
-For Rails applications using the default logger (`ActiveSupport::TaggedLogging`) or `lograge`, you can automatically enable trace correlation injection by setting the `rails` instrumentation configuration option `log_injection` to `true` or by setting environment variable `DD_LOGS_INJECTION=true`:
+For Rails applications using the default logger (`ActiveSupport::TaggedLogging`), `semantic_logger`, or `lograge`, you can automatically enable trace correlation injection by setting the `rails` instrumentation configuration option `log_injection` to `true` or by setting environment variable `DD_LOGS_INJECTION=true`:
 
 ```ruby
 # config/initializers/datadog.rb
@@ -2249,8 +2251,6 @@ Datadog.configure do |c|
 end
 ```
 
-_Note:_ For `lograge` users who have also defined `lograge.custom_options` in an `initializers/lograge.rb` configuration file, due to the order that Rails loads initializers (alphabetical), automatic trace correlation may not take effect, since `initializers/datadog.rb` would be overwritten by the `initializers/lograge.rb` initializer. To support automatic trace correlation with _existing_ `lograge.custom_options`, use the [Manual (Lograge)](#manual-lograge) configuration below.
-
 ##### Manual (Lograge)
 
 After [setting up Lograge in a Rails application](https://docs.datadoghq.com/logs/log_collection/ruby/), manually modify the `custom_options` block in your environment configuration file (e.g. `config/environments/production.rb`) to add the trace IDs.
@@ -2410,7 +2410,7 @@ The tracer and its integrations can produce some additional metrics that can pro
 To configure your application for metrics collection:
 
 1. [Configure your Datadog agent for StatsD](https://docs.datadoghq.com/developers/dogstatsd/#setup)
-2. Add `gem 'dogstatsd-ruby'` to your Gemfile
+2. Add `gem 'dogstatsd-ruby', '~> 5.2'` to your Gemfile
 
 #### For application runtime
 
@@ -2480,3 +2480,49 @@ However, additional instrumentation provided by Datadog can be activated alongsi
 | `OpenTracing::FORMAT_TEXT_MAP` | Yes        |                        |
 | `OpenTracing::FORMAT_RACK`     | Yes        | Because of the loss of resolution in the Rack format, please note that baggage items with names containing either upper case characters or `-` will be converted to lower case and `_` in a round-trip respectively. We recommend avoiding these characters or accommodating accordingly on the receiving end. |
 | `OpenTracing::FORMAT_BINARY`   | No         |                        |
+
+### Profiling
+
+*Currently available as BETA feature.*
+
+`ddtrace` can produce profiles that measure method-level application resource usage within production environments. These profiles can give insight into resources spent in Ruby code outside of existing trace instrumentation.
+
+**Setup**
+
+To get started with profiling, follow the [Profiler Getting Started Guide](https://docs.datadoghq.com/tracing/profiler/getting_started/?code-lang=ruby).
+
+#### Profiling Resque jobs
+
+When profiling [Resque](https://github.com/resque/resque) jobs, you should set the `RUN_AT_EXIT_HOOKS=1` option described in the [Resque](https://github.com/resque/resque/blob/v2.0.0/docs/HOOKS.md#worker-hooks) documentation.
+
+Without this flag, profiles for short-lived Resque jobs will not be available as Resque kills worker processes before they have a chance to submit this information.
+
+## Known issues and suggested configurations
+
+### Payload too large
+
+By default, Datadog limits the size of trace payloads to prevent memory overhead within instrumented applications. As a result, traces containing thousands of operations may not be sent to Datadog.
+
+If traces are missing, enable [debug mode](#tracer-settings) to check if messages containing `"Dropping trace. Payload too large"` are logged.
+
+Since debug mode is verbose, Datadog does not recommend leaving this enabled or enabling this in production. Disable it after confirming. You can inspect the [Datadog Agent logs](https://docs.datadoghq.com/agent/guide/agent-log-files/) for similar messages.
+
+If you have confirmed that traces are dropped due to large payloads, then enable the [partial_flush](#tracer-settings) setting to break down large traces into smaller chunks.
+
+### Stack level too deep
+
+Datadog tracing collects trace data by adding instrumentation into other common libraries (e.g. Rails, Rack, etc.) Some libraries provide APIs to add this instrumentation, but some do not. In order to add instrumentation into libraries lacking an instrumentation API, Datadog uses a technique called "monkey-patching" to modify the code of that library.
+
+In Ruby version 1.9.3 and earlier, "monkey-patching" often involved the use of [`alias_method`](https://ruby-doc.org/core-3.0.0/Module.html#method-i-alias_method), also known as *method rewriting*, to destructively replace existing Ruby methods. However, this practice would often create conflicts & errors if two libraries attempted to "rewrite" the same method. (e.g. two different APM packages trying to instrument the same method.)
+
+In Ruby 2.0, the [`Module#prepend`](https://ruby-doc.org/core-3.0.0/Module.html#method-i-prepend) feature was introduced. This feature avoids destructive method rewriting and allows multiple "monkey patches" on the same method. Consequently, it has become the safest, preferred means to "monkey patch" code.
+
+Datadog instrumentation almost exclusively uses the `Module#prepend` feature to add instrumentation non-destructively. However, some libraries (typically those supporting Ruby < 2.0) still use `alias_method` which can create conflicts with Datadog instrumentation, often resulting in `SystemStackError` or `stack level too deep` errors.
+
+As the implementation of `alias_method` exists within those libraries, Datadog generally cannot fix them. However, some libraries have known workarounds:
+
+* `rack-mini-profiler`: [Net::HTTP stack level too deep errors](https://github.com/MiniProfiler/rack-mini-profiler#nethttp-stack-level-too-deep-errors).
+
+For libraries without a known workaround, consider removing the library using `alias` or `Module#alias_method` or separating libraries into different environments for testing.
+
+For any further questions or to report an occurence of this issue, please [reach out to Datadog support](https://docs.datadoghq.com/help)