Ruby agent release notesRSS

v10.0.0

• Breaking Change: Remove support for Ruby 2.4 and 2.5

  Support for Ruby versions 2.4 and 2.5 has been removed. The new minimum required Ruby version is now 2.6. PR#3314

• Breaking Change: Removal of Cross Application Tracing (CAT)

  Previously, Cross Application Tracing (CAT) was deprecated in favor of Distributed Tracing. CAT functionality has now been removed. The configuration option cross_application_tracer.enabled has been removed. Public API methods NewRelic::Agent::External.process_request_metadata, NewRelic::Agent::External.get_response_metadata, NewRelic::Agent::Transaction::ExternalRequestSegment#process_response_metadata, NewRelic::Agent::Transaction::ExternalRequestSegment#get_request_metadata, and NewRelic::Agent::Transaction::ExternalRequestSegment#read_response_headers have also been removed. PR#3333

• Breaking Change: Rename ActiveJob metrics and segments

  ActiveJob metrics have been updated to include the job's class name for more specific reporting. This is a breaking change and may require updating custom dashboards or alerts. PR#3370 PR#3320

  • Old format: Ruby/ActiveJob/<QueueName>/<Method>
  • New format: Ruby/ActiveJob/<QueueName>/<ClassName>/<Method>

  In addition, segments created for Active Job enqueuing actions now include the job class.

  • Old format: ActiveJob/<QueueAdapter>/Queue/<Event>/Named/<QueueName>
  • New format: ActiveJob/<QueueAdapter>/Queue/<Event>/Named/<QueueName>/<ClassName>

• Breaking Change: Rename bin/newrelic command to bin/newrelic_rpm

  The executable file for the agent's CLI has been renamed from bin/newrelic to bin/newrelic_rpm. This change resolves a name collision with the standalone New Relic CLI tool. PR#3323

• Breaking Change: Remove the newrelic deployments CLI command

  The deprecated newrelic deployments CLI command has been removed. To track changes and deployments in New Relic, please see our guide to Change Tracking for a list of available options. PR#3299

• Breaking Change: Remove the NewRelic::Agent::SqlSampler#notice_sql method

  Users should call NewRelic::Agent::Datastores.notice_sql instead. PR#3338

• Breaking Change: Remove unused arguments from various NewRelic::Agent::Datastores APIs

  The following APIs from the NewRelic::Agent::Datastores class have had method arguments removed:

  • NewRelic::Agent::Datastores.notice_sql, previously had three positional arguments, query, scoped_metric and elapsed. Now, it only has query.
  • NewRelic::Agent::Datastores.notice_statement, previously had two positional arguments query and elapsed. Now it only has query.
  • NewRelic::Agent::Datastores.wrap requires a proc. Previously the proc received three arguments: the result of the yield, the most specific scoped metric name, and the elapsed time of the call. Now, it only receives one: the result of the yield.

  The values of the removed arguments are derived from the current segment at the time of the call. PR#3347

• Breaking Change: Remove experimental feature Configurable Security Policies (CSP)

  The experimental feature, Configurable Security Policies (CSP), is no longer supported and has been removed. PR#3292

• Breaking Change: Remove support for Puma versions < 3.9.0

  The minimum version of Puma now supported is 3.9.0 or higher. PR#3326

• Breaking Change: Improve configuration validation and coercion

  The internals used to coerce and validate the values provided for agent configuration are now more performant and more accurate.

  • Warning messages will now be logged to the newrelic_agent.log file when nil is provided as a config value for a setting that does not support it.
  • Integer values are permitted for Float configuration types
  • Float values are permitted for Integer configuration types
  • Fatal interruptions are prevented when a default value can be found to replace an invalid input value PR#3341

• Breaking Change: Replace 'default' option with 'adaptive' for distributed tracing remote parent samplers

  Previously, the default option for distributed_tracing.sampler.remote_parent_sampled and distributed_tracing.sampler.remote_parent_not_sampled was default, which used the pre-existing adaptive sampler. The default option has been renamed to adaptive. PR#3363

• Feature: Add logger as a dependency

  The logger gem is now listed as a dependency of the agent to ensure continued logging functionality and support for Ruby 4.0.0 and newer versions. PR#3293

• Feature: Add Active Support notification allowlist configuration option

  A new configuration option, instrumentation.active_support_notifications.active_support_events, allows users to define an allowlist of Active Support notifications event names for the agent to subscribe to. By default, the agent subscribes to all Active Support: Caching and Active Support: Messages events. PR#3327

• Feature: Use Ruby's built-in Gzip compression

  The agent now uses the built-in Zlib.gzip method from the Ruby standard library for compression, replacing the previous custom implementation. PR#3332

• Feature: Add argument validation for the NewRelic::Agent#record_custom_event API

  The NewRelic::Agent#record_custom_event API now raises an ArgumentError when an invalid event_type is provided. A valid event type must consist only of alphanumeric characters, underscores (_), colons (:), or spaces ( ). PR#3319

• Feature: Add root sampling configuration options

  You can now configure the sampling behavior for traces that originate within the current service using distributed_tracing.sampler.root. PR#3330

  There are four modes available:

  Mode	Description
  adaptive	Uses the existing adaptive sampler algorithm
  always_off	Marks 0% of root traces as sampled
  always_on	Marks 100% of root traces as sampled
  trace_id_ratio_based	Samples traces based on a ratio set in distributed_tracing.sampler.root.trace_id_ratio_based.ratio. The ratio must be float between 0.0 and 1.0

• Feature: Add Trace ID Ratio Based sampling options

  The agent can now sample traces using the OpenTelemetry Trace ID Ratio Based sampler algorithm. PR#3330 This samples traces based on a probability between 0.0 and 1.0 based on the trace ID.

  To use this option, you must first set your distributed tracing sampler configuration to trace_id_ratio_based and then set the corresponding distributed_tracing.sampler.*.trace_id_ratio_based.ratio sampler to a Float between 0.0 and 1.0.

  For example:

  distributed_tracing.sampler.remote_parent_sampled: 'trace_id_ratio_based'
    distributed_tracing.sampler.remote_parent_sampled.trace_id_ratio_based.ratio': 0.5

  Copy

  This configuration would sample approximately 50% of your traces for all traces where the remote parent is sampled.

  This option is available for:

  • distributed_tracing.sampler.root
  • distributed_tracing.sampler.remote_parent_sampled
  • distributed_tracing.sampler.remote_parent_not_sampled

• Feature: Add Entity GUID to Agent Control health check files

  When the agent is started within an Agent Control environment, a health check file is created at the configured file location for every agent process. This file now includes the guid of the entity related to the agent when available. PR#3371

• Bugfix: Resolve a NoMethodError in GCP utilization detection.

  The GCP metadata discovery logic will now gracefully handle nil or unexpected values, preventing service initialization crashes. PR#3388

v9.24.0

• Feature: Deprecation reminder for SqlSampler#notice_sql API

  The NewRelic::Agent::SqlSampler#notice_sql method is deprecated and will be removed in a future major version. Instead, users should call NewRelic::Agent::Datastores.notice_sql. PR#3345

• Feature: Deprecation notice for second and third arguments in Datastores.notice_sql API

  The second (scoped_metric) and third (elapsed) arguments in the NewRelic::Agent::Datastores.notice_sql method are deprecated. They have not been used by the method for some time. Instead, these values will be set based on the current segment when the API is called. PR#3345

• Feature: Deprecation notice for second argument in Datastores.notice_statement API

  The second (elapsed) argument in the NewRelic::Agent::Datastores.notice_statement method is deprecated. It has not been used by the method for some time. Instead, this value will be set based on the current segment when the API is called. PR#3346

• Feature: Deprecation notice for proc's second and third arguments in Datastores.wrap API

  The NewRelic::Agent::Datastores.wrap method is changing. In a future major version, proc will only accept a single argument, the result of the yield. The scoped metric name and elapsed arguments will be removed, as they are being removed from the Datastores.notice_sql method. The scoped metric name and elapsed values are derived from the current segment when the wrap yields. PR#3346

v9.23.0

• Feature: Add sidekiq.ignore_retry_errors configuration option

  A new configuration option, sidekiq.ignore_retry_errors, has been added to control if Sidekiq job retries are captured. Retry errors are captured by default, but now if sidekiq.ignore_retry_errors is set to true, the agent will ignore exceptions raised during Sidekiq's retry attempts and will only report the error if the job permanently fails. Thank you DonGiulio for recognizing this improvement and contributing a solution. PR#3317

• Feature: Deprecation notice for recording deployments using Capistrano

  Sending application deployment information using a Capistrano recipe is deprecated and will be removed in agent version 10.0.0. For recording deployments, please see our guide to Change Tracking for a list of available options.

• Feature: Use remote parent sampling configurations for decisions in more scenarios

  Previously, the distributed_tracing.sampler.remote_parent_sampled and distributed_tracing.sampler.remote_parent_not_sampled configuration options were used for the sampling decision only when the traceparent and tracestate headers were present. Now, these configuration options are applied in cases when the tracestate header is missing and when only the newrelic header is available. This change makes distributed trace sampling more consistent and predictable. PR#3306

v9.22.0

• Feature: One-step instrumentation for Kubernetes

  The Kubernetes APM auto-attach automatically instruments applications and manages agent upgrades within Kubernetes deployments. This feature has exited preview and is now generally avaliable. Learn more about Kubernetes auto-attach. PR#2635 PR#3287

• Feature: Deprecation notice for Ruby 2.4 and Ruby 2.5

  Ruby agent support for Ruby versions 2.4 and 2.5 is deprecated and will be removed in agent version 10.0.0. The new miniumum required Ruby version will become Ruby 2.6 and we will stop testing with Ruby 2.4 and 2.5. PR#3288

• Feature: Deprecation notice for the newrelic deployments command

  Support for recording deployments using the newrelic deployments command is now deprecated and will be removed in agent version 10.0.0.

  Going forward, there are a number of automated and manual ways ways to record changes in New Relic. Please see our guide to Change Tracking for a list of available options. PR#3262

• Feature: Deprecation reminder for cross application tracing

  Cross application tracing has been deprecated since major version 8.0.0 of the Ruby agent. We will remove support for it entirely in version 10.0.0 of the agent. PR#3288

• Feature: Ensure compatibility with Ruby 3.5 change to Method#source_location

  Updated the agent to correctly parse the return value of Method#source_location, which is changing in Ruby 3.5 from a two-element to a five-element array. This change maintains support for older Ruby versions while adding support for the future release. PR#3257

v9.21.0

• Feature: In Serverless APM mode, use event source name as transaction name prefix

  The agent will now use the event source name as a prefix for the transaction name in Serverless APM mode. This will help to better identify the source of the transaction in the New Relic UI. PR#3245

• Bugfix: Revert changed logic for how we track the thread the span starts in

  This change restores the previous behavior of tracking the thread the span starts in, addressing issues that arose from the updated logic. Thank you @david-zw-liu for bringing this to our attention. PR#3248

v9.20.0

• Feature: Add ECS Docker ID for Fargate

  Previously, the Ruby agent did not record the Docker ID when running in an AWS ECS Fargate environment. The Docker ID will now be recorded correctly. PR#3172

• Feature: Add NewRelic::Helper.version_satisfied?

  The agent has a new helper method to simplify version comparison. NewRelic::Helper.version_satisfied? accepts three arguments: a left-side version number, the comparison operator as a string, and a right-side version number. Our thanks go to @kekke-n for this contribution. PR#3182

• Feature: Add code.stacktrace attribute on datastore spans when duration exceeds configured threshold

  The agent will now add the code.stacktrace attribute to datastore spans when the duration exceeds the configured threshold. The threshold is configured using the transaction_tracer.stack_trace_threshold configuration option. PR#3220

• Feature: Consolidate "Unknown" constant values

  All references to the various capitalization styles for "Unknown" have been consolidated into two constants: NewRelic::UNKNOWN and NewRelic::UNKNOWN_LOWER. Thank you, @tsubasa1122, for your contribution! PR#3185

• Bugfix: Fix Brewfile source links

  Previously, the multiverse README's links to the Brewfile were broken. Our thanks go to @emmanuel-ferdman for submitting a PR to fix them! PR#3191

• Bugfix: Fix error when using HTTPX 1.5.0

  The agent previously encountered an error when using the new HTTPX version 1.5.0. This was due to a change in the way HTTPX stores the response. The agent has been updated to handle this change correctly, and no longer encounters an error when using HTTPX 1.5.0. PR#3203

• Bugfix: Bugfixes and improvements to debug level agent logs

  Improves the information logged at the debug level by the agent when the agent reads in a configuration source. PR#3221

• Bugfix: Fix risk of server-side forgery for Slack workflow script

  Internally, we keep track of gems that are released using a GitHub actions workflow that posts updates on Slack. [@odaysec] identified a way we could reduce the risk of server-side forgery for this workflow. Thank you! PR#3184

• Bugfix: Replace JSON.load calls with JSON.parse

  Generally, JSON.parse is seen as safer than JSON.load. Thank you, @odaysec, for bringing this to our attention! PR#3183 PR#3230

v9.19.0

• Feature: Add Thread ID as attribute to all spans

  The agent will now record the Thread ID as an attribute on each span. PR#3122

• Feature: Add support for W3C TraceContext Trace Flag

  Previously, the agent would not use the trace flag field of the traceparent header for sampling decisions. This could lead to fragmented traces in the UI. While the default behavior remains unchanged, two new configuration options, distributed_tracing.sampler.remote_parent_sampled and distributed_tracing.sampler.remote_parent_not_sampled, have been introduced to allow more control over the way sampling decisions are made. PR#3135

• Bugfix: Include request.uri in Transaction events by default

  The New Relic data dictionary expects Transaction events to have the request.uri attribute. The Ruby agent now fulfills this expectation. If you would like to exclude request.uri from Transaction events, you can do so by setting transaction_events.attributes.exclude to 'request.uri'. PR#3103

• Bugfix: Fix error in Active Job instrumentation when using perform_all_later

  Previously, when Active Job's perform_all_later method was called and the agent was running, a NoMethodError would be raised with the message undefined method 'queue_name' for nil. The error has been fixed and the name of the segment will reflect the first job in the queue. Our thanks goes to @tan-linx for bringing this to our attention and providing a fix. PR#3110

v9.18.0

• Feature: Add elasticsearch.capture_cluster_name configuration option

  A new configuration option, elasticsearch.capture_cluster_name, has been added to control capturing Elasticsearch cluster names. Cluster names are captured by default, but can now be disabled as needed. PR#3038

• Feature: Add support for sidekiq-delay_extensions

  Sidekiq delay extensions were removed from Sidekiq in 7.x and are now avaliable through the sidekiq-delay_extensions gem. Thanks to @sobrinho, the agent now has continued support for delay extensions.PR#3056

• Feature: Parallelize calls for vendor metadata

  Previously, the agent would make calls for vendor metadata in a serial fashion. This could lead to a delay in starting the agent. Now, the agent will make these calls in parallel, reducing the time it takes to start the agent. PR#3094

• Bugfix: Prevent a nil segment from causing errors in Net::HTTP instrumentation

  When using JRuby, a race condition can happen that causes the segment creation to fail and return nil. This would cause an error to occur when methods were later called on the nil segment. These methods will no longer be called if the segment is nil, preventing that error from occurring. PR#3046

• Bugfix: JRuby multithreading improvements

  Added some additional nil checks and mutexes to prevent issues when using the agent on JRuby with multiple threads. Thanks to @NC-piercej for bringing this to our attention Issue#3021 PR#3053

• Bugfix: Stop reporting rescued Sidekiq::OverLimit exceptions

  When Sidekiq's concurrent rate limiters encounter an OverLimit exception, Sidekiq typically handles this by re-enqueuing the job. Previously, all occurrences of Sidekiq::OverLimit were logged as errors in New Relic, even when Sidekiq's middleware resolved the exception. New Relic will no longer report errors that are handled by Sidekiq's own middleware. Thanks to @97jaz for reporting this issue. Issue#3037 PR#3047

• Bugfix: Protect against nil agents or health checks

  In some cases the agent or health checks may be nil when they are called. Safe navigation operators have been added for protection on those occasions. PR#3049

• Bugfix: Ignore Solid Queue ThreadError: queue empty error message by default

  When using the solid_queue gem, the agent previously generated excessive warn-level logs when the queue was empty. The agent now ignores queue empty error messages of the ThreadError class by default. This behavior can be adjusted using the error_collector.ignore_messages configuration option. PR#3060

• Bugfix: Refactor URI host handling to accommodate downcasing frozen strings

  When URI host string was frozen, a FrozenError would be raised when the agent attempted to downcase the host as part of its data normalization process. Now, the update is friendly for frozen strings. Thank you @pedrol3001 for your contribution! PR#3097

v9.17.0

• Feature: Support Ruby 3.4.0

  The agent now supports Ruby 3.4.0. We've made incremental changes throughout the preview stage to reach compatibility. This release includes an update to the Thread Profiler for compatibility with Ruby 3.4.0's new backtrace format. Issue#2992 PR#2997

• Feature: Add instrumentation for aws-sdk-firehose

  The agent now has instrumentation for the aws-sdk-firehose gem. PR#2973

• Feature: Kubernetes APM auto-attach - new agent version precedent

  Previously, when a customer installed the Ruby agent via Kubernetes APM auto-attach and also had the Ruby agent listed in their Gemfile, the agent version in Gemfile would take precedence. Now, the agent version installed by auto-attach takes priority. PR#3018

• Feature: Add health checks when the agent runs within Agent Control

  When the agent is started within an Agent Control environment, a health check file will be created at the configured file location for every agent process. By default, this location is: '/newrelic/apm/health'. The health check files will be updated at the configured frequency, which defaults to every five seconds. PR#2995

• Feature: Add Redshift as recognized ActiveRecord adapter

  When the agent does not recognize an ActiveRecord adapter, the host, port, and database name information is not added to the datastore span. Redshift will now be treated like PostgreSQL, and the agent will save the host, port, and database name on the span. PR#3032

• Feature: Add instrumentation for aws-sdk-kinesis

  The agent now has instrumentation for the aws-sdk-kinesis gem. It will record message broker segments for get_records, put_record, and put_records operations. All other operations will record standard segments. PR#2974

• Bugfix: Stop emitting inaccurate debug-level log about deprecated configuration options

  In the previous major release, we dropped support for many disable_library_name configuration options in favor of instrumentation.library_name. Previously, a DEBUG level log warning appeared whenever disable_* options were set to true, even for libraries (e.g. Action Dispatch) without equivalent instrumentation.* options:

>DEBUG : [DEPRECATED] configuration disable_library_name for library_name will be removed in the next major release. Use instrumentation. library_name with one of ["auto", "disabled", "prepend", "chain"]

This inaccurate warning has been removed. If you are disabling instrumentation using instrumentation.library_name: disabled or NEW_RELIC_INSTRUMENTATION_LIBRARY_NAME=disabled, please verify the option exists by consulting our configuration documentation. If the option does not exist, check the 'Disabling' section to see if there is a related option. We apologize for the confusion. PR#3005

• Bugfix: Do not attempt to decorate logs with nil messages

  The agent no longer attempts to add New Relic linking metadata to logs with nil messages. Thank you, @arlando for bringing this to our attention! Issue#2985 PR#2986

• Bugfix: Stop renaming final Grape segment

  Previously, the agent renamed the final segment in Grape transactions to "Middleware/Grape/#{class_name}/call". This was a part of an old instrumentation pattern that is no longer relevant. Many thanks to @seriousdev-gh for bringing this issue to our attention and along with a great reproduction and suggested fix. PR#2987.