Ruby agent release notesRSS

Feature: Add instrumentation for ruby-openai

Instrumentation has been added for the ruby-openai gem, supporting versions 3.4.0 and higher (PR#2442). While ruby-openai instrumentation is enabled by default, the configuration option ai_monitoring.enabled is disabled by default and controls all AI monitoring. ai_monitoring.enabled must be set to true in order to receive ruby-openai instrumentation. High-Security Mode must be disabled in order to receive AI monitoring.

Calls to embedding and chat completion endpoints are automatically traced. These events can be enhanced with the introduction of two new APIs. Custom attributes can also be added to LLM events using the API NewRelic::Agent.add_custom_attributes, but they must be prefixed with llm.. For example, NewRelic::Agent.add_custom_attributes({'llm.user_id': user_id}).

Feature: Add AI monitoring APIs

This version introduces two new APIs that allow users to record additional information on LLM events:

• NewRelic::Agent.record_llm_feedback_event - Records user feedback events.
• NewRelic::Agent.set_llm_token_count_callback - Sets a callback proc for calculating token_count attributes for embedding and chat completion message events.

Visit RubyDoc for more information on each of these APIs.

Feature: Store tracer state on thread-level

A new configuration option, thread_local_tracer_state, stores New Relic's tracer state on the thread-level, as opposed to the default fiber-level storage. This configuration is turned off by default. Our thanks go to community member @markiz who contributed the idea, code, configuration option, and tests for this new feature! PR#2475.

Bugfix: Harden the browser agent insertion logic

With Issue#2462, community member @miry explained that it was possible for an HTTP response headers hash to have symbols for values. Not only would these symbols prevent the inclusion of the New Relic browser agent tag in the response body, but more importantly they would cause an exception that would bubble up to the monitored web application itself. With PR#2465 symbol based values are now supported and all other potential future exceptions are now handled. Additionally, the refactor to support symbols has been shown through benchmarking to give the processing of string and mixed type hashes a slight speed boost too.

Bugfix: Prevent Exception in Active Support Broadcast logger instrumentation

Previously, in certain situations the agent could cause an exception to be raised when attempting to interact with a broadcast log event. This has been fixed. Thanks to @nathan-appere for reporting this issue and providing a fix! PR#2510

Bugfix: Stop suppressing ViewComponent errors

Previously, the agent suppressed ViewComponent render errors. The agent now reports these errors and allows them to raise. Thank you @mjacobus for reporting this bug and providing a fix! PR#2410

Bugfix: Enforce maximum size limits for custom event attributes

Previously, the agent would allow custom event attributes to be any size. This would lead to the New Relic backend dropping attributes larger than the maximum size. Now, the agent will truncate custom event attribute values to 4095 characters, attribute names to 255 characters, and the total count of attributes to 64. PR#2401

Feature: ViewComponent instrumentation

ViewComponent is a now an instrumented library. PR#2367

Feature: Use root path to access Elasticsearch cluster name

Previously, the agent used the cluster health endpoint (/_cluster/health) to access the cluster name. However, this has been found to make startup unstable for large clusters. Now, the agent uses the more performant root endpoint (/).

Our thanks go to @erikkessler1, @gremerritt, and @joshbranham for reporting the issue, suggesting solutions, and testing them. Issue#2360 PR#2377

Feature: Remove base64 dependency, use direct calls to String methods

In version 9.6.0, the agent required the Ruby base64 gem as a depdendency to prepare for deprecation warnings in Ruby 3.3 and the gem's removal from the Ruby standard libraries in 3.4. Including base64 as a dependency has caused problems with version resolution in some environments.

To resolve this, the agent now directly calls the String methods used in the base64 library in the new NewRelic::Base64 module.

Thank you, @Earlopain, for submitting this change. PR#2378

Feature: Add Falcon support

The agent now supports the web server Falcon. PR#2383

Feature: Remove spans with name Ruby/Thread and Ruby/Fiber

Due to the lack of helpful information and the confusion commonly caused by the spans named Ruby/Thread and Ruby/Fiber, these spans have been removed. However, the agents ability to monitor instrumented code running in a thread or fiber will remain unchanged. PR#2389

Feature: Add instrumentation for Async::HTTP

The agent will now record spans for Async::HTTP requests. Versions 0.59.0 and above of the async-http gem are supported. PR#2272

Feature: Add instrumentation for Ethon

Instrumentation has been added for the Ethon HTTP client gem. Versions 0.12.0 and above are supported. The agent will now record external request segments for invocations of Ethon::Easy#perform and Ethon::Multi#perform. NOTE: The Typhoeus gem is maintained by the same team that maintains Ethon and depends on Ethon for its functionality. To prevent duplicate reporting for each HTTP request, the Ethon instrumentation will be disabled when Typhoeus is detected. PR#2260

Feature: Add instrumentation for HTTPX

The agent now offers instrumentation for the HTTP client HTTPX, provided the gem is at version 1.0.0 or above. PR#2278

Feature: Prevent the agent from starting in "rails" commands in Rails 7

Previously, the agent ignored many Rails commands by default, such as rails routes, using Rake-specific logic. This was accomplished by setting these commands as default values for the config option autostart.denylisted_rake_tasks. However, Rails 7 no longer uses Rake for these commands, causing the agent to start running and attempting to record data when running these commands. The commands have now been added to the default value for the config option autostart.denylisted_constants, which will allow the agent to recognize these commands correctly in Rails 7 and prevent the agent from starting during ignored tasks. Note that the agent will continue to start-up when the rails server and rails runner commands are invoked. PR#2239

Feature: Glean Docker container ID for cgroups v2-based containers

Previously, the agent was only capable of determining a host Docker container's ID if the container was based on cgroups v1. Now, containers based on cgroups v2 will also have their container IDs reported to New Relic. PR#2229.

Feature: Update events with additional synthetics attributes when available

The agent will now record additional synthetics attributes on synthetics events if these attributes are available. PR#2203

Feature: Declare a gem dependency on the Ruby Base 64 gem 'base64'

For compatibility with Ruby 3.4 and to silence compatibility warnings present in Ruby 3.3, declare a dependency on the base64 gem. The New Relic Ruby agent uses the native Ruby base64 gem for Base 64 encoding/decoding. The agent is joined by Ruby on Rails (rails/rails@3e52adf) and others in making this change in preparation for Ruby 3.3/3.4. PR#2238

Feature: Add Roda support for the newrelic_ignore* family of methods

The agent can now selectively disable instrumentation for particular requests within Roda applications. Supported methods include:

• newrelic_ignore: ignore a given route.
• newrelic_ignore_apdex: exclude a given route from consideration in overall Apdex calculations.
• newrelic_ignore_enduser: prevent automatic injection of the page load timing JavaScript when a route is rendered.

For more information, see Roda Instrumentation. PR#2267

Feature: Add additional span attributes for OpenTelemetry compatibility

For improved compatibility with OpenTelemetry's semantic conventions, the agent's datastore (for databases) and external request (for HTTP clients) segments have been updated with additional attributes.

Datastore segments now offer 3 additional attributes:

• db.system: The database system. For Ruby we use the database adapter name here.
• server.address: The database host.
• server.port: The database port.

External request segments now offer 3 additional attributes:

• http.request.method: The HTTP method (ex: 'GET')
• server.address: The target host.
• server.port: The target port.

For maximum backwards compatibility, no existing attributes have been renamed or removed. PR#2283

Bugfix: Stop sending duplicate log events for Rails 7.1 users

Rails 7.1 introduced the public API ActiveSupport::BroadcastLogger. This logger replaces a private API, ActiveSupport::Logger.broadcast. In Rails versions below 7.1, the agent uses the broadcast method to stop duplicate logs from being recoded by broadcasted loggers. Now, we've updated the code to provide a similar duplication fix for the ActiveSupport::BroadcastLogger class. PR#2252

Bugfix: Resolve Sidekiq 8.0 error handler deprecation warning

Sidekiq 8.0 will require procs passed to the error handler to include three arguments: error, context, and config. Users running sidekiq/main would receive a deprecation warning with this change any time an error was raised within a job. Thank you, @fukayatsu for your proactive fix! PR#2261

Community: Resolve technical debt

We also received some great contributions from community members to resolve some outstanding technical debt issues. Thank you for your contributions!

• Add and Replace SLASH and ROOT constants: PR#2256 chahmedejaz
• Remove pry as a dev dependency: PR#2665, PR#2273 AlajeBash
• Replace "start up" with "start-up": PR#2249 chahmedejaz
• Remove unused variables in test suites: PR#2250

Feature: Add Stripe instrumentation

Stripe calls are now automatically instrumented. Additionally, new :'stripe.user_data.include' and :'stripe.user_data.exclude' configuration options permit capturing custom user_data key-value pairs that can be stored in Stripe events. No user_data key-value pairs are captured by default. The agent currently supports Stripe versions 5.38.0+. PR#2180

Feature: Report transaction HTTP status codes when middleware instrumentation is disabled

Previously, when disable_middleware_instrumentation was set to true, the agent would not record the value of the response code or content type on the transaction. This was due to the possibility that a middleware could alter the response, which would not be captured by the agent when the middleware instrumentation was disabled. However, based on customer feedback, the agent will now report the HTTP status code and content type on a transaction when middleware instrumentation is disabled. PR#2175

Feature: Permit capturing only certain Sidekiq job arguments

New :'sidekiq.args.include' and :'sidekiq.args.exclude' configuration options have been introduced to permit fine grained control over which Sidekiq job arguments (args) are reported to New Relic. By default, no Sidekiq args are reported. To report any Sidekiq options, the :'attributes.include' array must include the string 'jobs.sidekiq.args.*'. With that string in place, all arguments will be reported unless one or more of the new include/exclude options are used. The :'sidekiq.args.include' option can be set to an array of strings. Each of those strings will be passed to Regexp.new and collectively serve as an allowlist for desired args. For job arguments that are hashes, if a hash's key matches one of the include patterns, then both the key and its corresponding value will be included. For scalar arguments, the string representation of the scalar will need to match one of the include patterns to be captured. The :'sidekiq.args.exclude' option works similarly. It can be set to an array of strings that will each be passed to Regexp.new to create patterns. These patterns will collectively serve as a denylist for unwanted job args. Any hash key, hash value, or scalar that matches an exclude pattern will be excluded (not sent to New Relic). PR#2177

newrelic.yml examples:

Any string in the :'sidekiq.args.include' or :'sidekiq.args.exclude' arrays will be turned into a regular expression. Knowledge of Ruby regular expression support can be leveraged but is not required. If regular expression syntax is not used, inexact matches will be performed and the string "Fortune" will match both "Fortune 500" and "Fortune and Glory". For exact matches, use regular expression anchors.

# Include any argument whose string representation matches either "apple" or "banana"
# The "apple" pattern will match both "green apple" and "red apple"
sidekiq.args.include:
  - apple
  - banana

# Exclude any arguments that match either "grape", "orange", or "pear"
sidekiq.args.exclude:
  - grape
  - orange
  - pear

# Exclude any argument that is a 9 digit number
sidekiq.args.exclude:
  - '\d{9}'

# Include anything that starts with "blue" but exclude anything that ends in "green"
sidekiq.args.include
  - '^blue'

sidekiq.args.exclude
  - 'green$'

Bugfix: Update Elasticsearch datastore instance metric to use port instead of path

Previously, the Elasticsearch datastore instance metric (Datastore/instance/Elasticsearch/<host>/*) used the path as the final value. This caused a metrics grouping issue for some users, as every document ID created a unique metric. Now, the datastore instance metric has been updated to use the port as the final value. This also has the benefit of being more accurate for datastore instance metrics, as this port is directly associated with the already listed host.

Bugfix: Resolve inverted logic of NewRelic::Rack::AgentHooks.needed?

Previously, NewRelic::Rack::AgentHooks.needed? incorrectly used inverted logic. This has now been resolved, allowing AgentHooks to be installed when disable_middleware_instrumentation is set to true. PR#2175

Bugfix: Resolve NoMethodError

Ruby agent 9.4.1 attempted to fix a NoMethodError introduced in 9.4.0. A missing require prevented a method from scoping appropriately and has now been added. Thanks to @spickermann and @ColinOrr for working with us to get this resolved. PR#2167

Bugfix: Resolve NoMethodError

Ruby agent 9.4.0 introduced Roda instrumentation, which caused a NoMethodError to be raised when attempting to name a Roda transaction. This has been fixed. Thanks to @spickermann for reporting this issue. PR#2167

Feature: Add Roda instrumentation

Roda is a now an instrumented framework. The agent currently supports Roda versions 3.19.0+. PR#2144

Feature: New allow_all_headers configuration option

A new allow_all_headers configuration option brings parity with the Node.js agent to capture all HTTP request headers.

This configuration option:

• Defaults to false
• Is not compatible with high-security mode
• Requires Rack version 2 or higher (as does Ruby on Rails version 5 and above)
• Respects all existing behavior for the attributes.include and attributes.exclude configuration options
• Captures the additional headers as attributes prefixed with request.headers.

This work was done in response to a feature request submitted by community member @jamesarosen. Thank you very much, @jamesarosen! Issue#1029

Feature: Improved error tracking transaction linking

Errors tracked and sent to the New Relic errors inbox will now be associated with a transaction id to enable improved UI/UX associations between transactions and errors. PR#2035

Feature: Use Net::HTTP native timeout logic

In line with current Ruby best practices, make use of Net::HTTP's own timeout logic and avoid the use of Timeout.timeout() when possible. The agent's data transmissions and cloud provider detection routines have been updated accordingly. PR#2147

Bugfix: Fix NewRelic::Agent.require_test_helper

Version 9.3.0 of the agent made a change to the files distributed with the gem. This change unintentionally broke the NewRelic::Agent.require_test_helper API by removing the test/agent_helper.rb file. The file has been added back to the gem. This change also removes the lib/new_relic/build.rb file from the list because it is no longer created with our current release process.

Our thanks go to @ajesler for reporting this issue and writing a test for the bug. Issue#2113, PR#2115, Issue#2117, PR#2118

Source Documentation: update the Rack spec URL

Community member @olleolleolle noticed that our source code was referencing a now defunct URL for the Rack specification and submitted PR#2121 to update it. He also provided a terrific recommendation that we automate the checking of links to proactively catch defunct ones in future. Thanks, @olleolleolle!