Ruby agent release notesRSS

Rails 4.2 ActiveJob support

A new version of Rails is coming! One of the highlight features is ActiveJob, a framework for interacting with background job processors. This release of the Ruby agent adds instrumentation to give you insight into ActiveJob, whether you're just testing it out or running it for real.

Metrics are recorded around enqueuing ActiveJobs, and background transactions are started for any ActiveJob performed where the agent didn't already provide specific instrumentation (such as DelayedJob, Resque and Sidekiq).

Since Rails 4.2 is still in beta we'd love to hear any feedback on this instrumentation so it'll be rock solid for the general release!

Ruby 2.2.0-preview1 updates

Ruby 2.2.0 is on its way later in the year, and the Ruby agent is ready for it. Updates to the GC stats and various other small changes have already been applied, and our automated tests are running against 2.2.0 so the agent will be ready on release day.

Ignoring transactions by URL

While you could always ignore transactions by controller and action, the Ruby agent previously lacked a way to ignore by specific URLs or patterns without code changes. This release adds the config setting, rules.ignore_url_regexes to ignore specific transactions based on the request URL as well. For more information, see the documentation at: https://docs.newrelic.com/docs/agents/ruby-agent/installation-configurat...

Better dependency detection in non-Rack applications

The Ruby agent runs dependency detection at key points in the Rack and Rails lifecycle, but non-Rails apps could occasionally miss out instrumenting late loaded libraries. The agent now runs an additional dependency detection during manual_start to more seamlessly install instrumentation in any app.

Excluding /newrelic routes from developer mode

Recent changes to track time in middleware resulted in New Relic's developer mode capturing its own page views in the list. This has been fixed. Thanks to Ignatius Reza Lesmana for the report!

Spikes in external time

Timeouts during certain external HTTP requests could result in incorrect large spikes in the time recorded by the agent. This has been fixed.

Recognize browser_monitoring.auto_instrument setting in non-Rails apps

The browser_monitoring.auto_instrument config setting disables auto-injection of JavaScript into your pages, but was not properly obeyed in Sinatra and other non-Rails contexts. This has been fixed.

Failures to gather CPU thread time on JRuby

JRuby running on certain JVM's and operating systems (FreeBSD in particular) did not always support the method being used to gather CPU burn metrics. This would result in a failure during those transactions. This has been fixed.

Fix for rare race condition in Resque instrumentation

A race condition in the agent's Resque instrumentation that could cause rare Resque job failures in high-throughput Resque setups has been fixed. This bug would manifest as an exception with the following error message: "RuntimeError: can't add a new key into hash during iteration" and a backtrace leading through the PipeChannelManager class in the agent.

Per-dyno data on Heroku

When running on Heroku, data from the agent can now be broken out by dyno name, allowing you to more easily see what's happening on a per-dyno level. Dynos on Heroku are now treated in the same way that distinct hosts on other platforms work.

By default, 'scheduler' and 'run' dyno names will be aggregated into scheduler.* and run.* to avoid unbounded growth in the number of reported hostnames.

Read more about this feature on our Heroku docs page: https://docs.newrelic.com/docs/agents/ruby-agent/miscellaneous/ruby-agen...

HTTP response codes in Insights events

The Ruby agent will now capture HTTP response codes from Rack applications (including Rails and Sinatra apps) and include them under the httpResponseCode attribute on events sent to Insights.

Stricter limits on memory usage of metrics and SQL traces

The agent now imposes stricter limits on the number of distinct SQL traces and metrics that it will buffer in memory at any point in time, leading to more predictable memory consumption even in exceptional circumstances.

Improved reliability of thread profiling

Several issues that would previously have prevented the successful completion and transmission of thread profiles to New Relic's servers have been fixed.

These issues were related to the use of recursion in processing thread profiles, and have been addressed by both limiting the maximum depth of the backtraces recorded in thread profiles, and eliminating the agent's use of recursion in processing profile data.

Allow tracing Rails view helpers with add_method_tracer

Previously, attempting to trace a Rails view helper method using add_method_tracer on the view helper module would lead to a NoMethodError when the traced method was called (undefined method `trace_execution_scoped'). This has been fixed.

This issue was an instance of the Ruby 'dynamic module inclusion' or 'double inclusion' problem. Usage of add_method_tracer now no longer relies upon the target class having actually picked up the trace_execution_scoped method from the NewRelic::Agent::MethodTracer module.

Improved performance of queue time parsing

The number of objects allocated while parsing the frontend timestamps on incoming HTTP requests has been significantly reduced.

Thanks to Aleksei Magusev for the contribution!

Allow agent to use alternate certificate stores

When connecting via SSL to New Relic, the Ruby agent verifies its connection via a certificate bundle shipped with the agent. This caused problems with certain proxy configurations, so the ca_bundle_path setting in newrelic.yml can now override the default cert bundle location.

For more information, see custom SSL certificates

Rails 4.2 beta in tests

Although still in beta, a new version of Rails is on its way! We're already running our automated test suites against the beta to ensure New Relic is ready the day the next Rails is released.

ActiveRecord 4 cached queries fix

Queries against the ActiveRecord 4.x query cache were incorrectly being counted as database time by the agent.

Fix for error in newrelic.yml loading

If your application used a RAILS_ENV value that was not listed in newrelic.yml recent agent versions would return a NameError rather than a helpful message. This has been fixed. Thanks Oleksiy Kovyrin for the patch!

Fix to prevent proxy credentials transmission

This update prevents proxy credentials set in the agent config file from being transmitted to New Relic.

Added API for ignoring transactions

This release adds three new API calls for ignoring transactions:

• NewRelic::Agent.ignore_transaction
• NewRelic::Agent.ignore_apdex
• NewRelic::Agent.ignore_enduser

ignore_transaction ignores a transaction completely: nothing about it will be reported to New Relic. ignore_apdex ignores only the Apdex metric for a single transaction. ignore_enduser disables Javascript injection for browser monitoring for the current transaction.

These methods differ from the existing newrelic_ignore_* method in that they may be called during a transaction based on some dynamic runtime criteria, as opposed to at the class level on startup.

For more information, see ignoring specific transactions.

Improved SQL obfuscation

SQL queries containing string literals ending in backslash \ characters were not correctly obfuscated by the Ruby agent prior to transmission to New Relic. In addition, SQL comments were left un-obfuscated. This has been fixed, and the test coverage for SQL obfuscation has been improved.

newrelic_ignore* methods now work when called in a superclass

The newrelic_ignore* family of methods previously did not apply to subclasses of the class from which it was called, meaning that Rails controllers inheriting from a single base class where newrelic_ignore had been called would not be ignored. This has been fixed.

Fix for rare crashes in Rack::Request#params on Sinatra apps

Certain kinds of malformed HTTP requests previously caused unhandled exceptions in the Ruby agent's Sinatra instrumentation, in the Rack::Request#params method. This has been fixed.

Improved handling for rare errors caused by timeouts in Excon requests

In some rare cases, the agent would emit a warning message in its log file and abort instrumentation of a transaction if a timeout occurred during an Excon request initiated from within that transaction. This has been fixed.

Improved behavior when the agent is misconfigured

When the agent is misconfigured by attempting to shut it down without it ever having been started, or by attempting to disable instrumentation after instrumentation has already been installed, the agent will no longer raise an exception, but will instead log an error to its log file.

Fix for ignore_error_filter not working in some configurations

The ignore_error_filter method allows you to specify a block to be evaluated in order to determine whether a given error should be ignored by the agent. If the agent was initially disabled, and then later enabled with a call to manual_start, the ignore_error_filter would not work. This has been fixed.

Fix for Capistrano 3 ignoring newrelic_revision

New Relic's Capistrano recipes support passing parameters to control the values recorded with deployments, but user-provided :newrelic_revision values were incorrectly overwritten. This has been fixed.

Agent errors logged with ruby-prof in production

If the ruby-prof gem was available in an environment without New Relic's developer mode enabled, the agent would generate errors to its log. This has been fixed.

Tighter requirements on naming for configuration environment variables

The agent would previously assume any environment variable containing NEWRELIC was a configuration setting. It now looks for this string as a prefix only.

Thanks to Chad Woolley for the contribution!

Ruby 1.8.7 users: upgrade or add JSON gem now

Ruby 1.8.7 is end-of-lifed, and not receiving security updates, so we strongly encourage all users with apps on 1.8.7 to upgrade.

If you're not able to upgrade yet, be aware that a coming release of the Ruby agent will require users of Ruby 1.8.7 to have the 'json' gem available within their applications in order to continue sending data to New Relic.

Support for new Cross Application Trace view

This release enhances cross application tracing with a visualization of the cross application calls that a specific Transaction Trace is involved in. The new visualization helps you spot bottlenecks in external services within Transaction Traces and gives you an end-to-end understanding of how the transaction trace is used by other applications and services. This leads to faster problem diagnosis and better collaboration across teams. All agents involved in the cross application communication must be upgraded to see the complete graph. You can view cross application traces from in the Transaction Trace drill-down.

High-security mode v2

The Ruby agent now supports v2 of New Relic's high-security mode. To enable it, you must add 'high_security: true' to your newrelic.yml file, and enable high-security mode through the New Relic web interface. The local agent setting must be in agreement with the server-side setting, or the agent will shut down and no data will be collected.

Customers who already had the server-side high-security mode setting enabled must add 'high_security: true' to their agent configuration files when upgrading to this release.

For details on high-security mode, see: http://docs.newrelic.com/docs/accounts-partnerships/accounts/security/hi...

Improved memcached instrumentation

More accurate instrumentation for the 'cas' command when using version 1.8.0 or higher of the memcached gem. Previous versions of the agent would count all time spent in the block given to 'cas' as memcache time, but 1.8.0 or higher allows us to more accurately measure just the time spent talking to memcache.

Many thanks to Francis Bogsanyi for contributing this change!

Improved support for Rails apps launched from outside the app root directory

The Ruby agent attempts to resolve the location of its configuration file at runtime relative to the directory that the host process is started from.

In cases where the host process was started from outside of the application's root directory (for example, if the process is started from from '/'), it will now also attempt to locate its configuration file based on the value of Rails.root for Rails applications.

Better compatibility with ActionController::Live

Browser Application Monitoring auto-injection can cause request failures under certain circumstances when used with ActionController::Live, so the agent will now automatically detect usage of ActionController::Live, and not attempt auto-injection for those requests (even if auto-instrumentation is otherwise enabled).

Many thanks to Rodrigo Rosenfeld Rosas for help diagnosing this issue!

Fix for occasional spikes in external services time

Certain kinds of failures during HTTP requests made by an application could have previously resulted in the Ruby agent reporting erroneously large amounts of time spent in outgoing HTTP requests. This issue manifested most obviously in spikes on the 'Web external' band on the main overview graph. This issue has now been fixed.

Fix 'rake newrelic:install' for Rails 4 applications

The newrelic:install rake task was previously not working for Rails 4 applications and has been fixed.

Thanks to Murahashi Sanemat Kenichi for contributing this fix!

Rack middleware instrumentation

The Ruby agent now automatically instruments Rack middlewares!

This means that the agent can now give you a more complete picture of your application's response time, including time spent in middleware. It also means that requests which previously weren't captured by the agent because they never made it to the bottom of your middleware stack (usually a Rails or Sinatra application) will now be captured.

After installing this version of the Ruby agent, you'll see a new 'Middleware' band on your application's overview graph, and individual middlewares will appear in transaction breakdown charts and transaction traces.

The agent can instrument middlewares that are added from a config.ru file via Rack::Builder, or via Rails' middleware stack in Rails 3.0+.

This instrumentation may be disabled with the disable_middlware_instrumentation configuration setting.

For more details, see the documentation for this feature:

• http://docs.newrelic.com/docs/ruby/rack-middlewares
• http://docs.newrelic.com/docs/ruby/middleware-upgrade-changes

Capistrano 3.x support

Recording application deployments using Capistrano 3.x is now supported.

Many thanks to Jennifer Page for the contribution!

Better support for Sidekiq's Delayed extensions

Sidekiq jobs executed via the Delayed extensions (e.g. the #delay method) will now be named after the actual class that #delay was invoked against, and will have their job arguments correctly captured if the sidekiq.capture_params configuration setting is enabled.

Many thanks to printercu for the contribution!

Improved Apdex calculation with ignored error classes

Previously, a transaction resulting in an exception that bubbled up to the top level would always be counted as failing for the purposes of Apdex calculations (unless the transaction name was ignored entirely). Now, exceptions whose classes have been ignored by the error_collector.ignore_errors configuration setting will not cause a transaction to be automatically counted as failing.

Allow URIs that are not parseable by stdlib's URI if addressable is present

There are some URIs that are valid by RFC 3986, but not parseable by Ruby's stdlib URI class. The Ruby agent will now attempt to use the addressable gem to parse URIs if it is present, allowing requests against these problematic URIs to be instrumented.

Many thanks to Craig R Webster and Amir Yalon for their help with this issue!

More robust error collection from Resque processes

Previously, traced errors where the exception class was defined in the Resque worker but not in the Resque master process would not be correctly handled by the agent. This has been fixed.

Allow Sinatra apps to set the New Relic environment without setting RACK_ENV

The NEW_RELIC_ENV environment variable may now be used to specify the environment the agent should use from its configuration file, independently of RACK_ENV.

Many thanks to Mario Izquierdo for the contribution!

Better error handling in Browser Application Monitoring injection

The agent middleware that injects the JavaScript code necessary for Browser Application Monitoring now does a better job of catching errors that might occur during the injection process.

Allow disabling of Net::HTTP instrumentation

Most instrumentation in the Ruby agent can be disabled easily via a configuration setting. Our Net::HTTP instrumentation was previously an exception, but now it can also be disabled with the disable_net_http configuration setting.

Make Rails constant presence check more defensive

The Ruby agent now guards against the (rare) case where an application has a Rails constant defined, but no Rails::VERSION constant (because Rails is not actually present).

Many thanks to Vladimir Kiselev for the contribution!

Better handling for Rack applications implemented as middlewares

When using a Sinatra application as a middleware around another app (for example, a Rails app), or manually instrumenting a Rack middleware wrapped around another application, the agent would previously generate two separate transaction names in the New Relic UI (one for the middleware, and one for the inner application).

As of this release, the agent will instead unify these two parts into a single transaction in the UI. The unified name will be the name assigned to the inner-most traced transaction by default. Calls to NewRelic::Agent.set_transaction_name will continue to override the default names assigned by the agent's instrumentation code.

This change also makes it possible to run X-Ray sessions against transactions of the 'inner' application in cases where one instrumented app is wrapped in another that's implemented as a middleware.

Support for mongo-1.10.0

The Ruby agent now instruments version 1.10.0 of the mongo gem (versions 1.8.x and 1.9.x were already supported, and continue to be).

Allow setting configuration file path via an option to manual_start

Previously, passing the :config_path option to NewRelic::Agent.manual_start would not actually affect the location that the agent would use to look for its configuration file. This has been fixed, and the log messages emitted when a configuration file is not found should now be more helpful.

Better support for forking and daemonizing dispatchers (e.g. Puma, Unicorn)

The agent should now work out-of-the box with no special configuration on servers that fork or daemonize themselves (such as Unicorn or Puma in some configurations). The agent's background thread will be automatically restarted after the first transaction processed within each child process.

This change means it's no longer necessary to set the 'restart_thread_in_children setting' in your agent configuration file if you were doing so previously.

Rails 4.1 support

Rails 4.1 has shipped, and the Ruby agent is ready for it! We've been running our test suites against the release candidates with no significant issues, so we're happy to announce full compatibility with this new release of Rails.

Ruby VM measurements

The Ruby agent now records more detailed information about the performance and behavior of the Ruby VM, mainly focused around Ruby's garbage collector. This information is exposed on the new 'Ruby VM' tab in the UI. For details about what is recorded, see:

http://docs.newrelic.com/docs/ruby/ruby-vm-stats

Separate in-transaction GC timings for web and background processes

Previously, an application with GC instrumentation enabled, and both web and background processes reporting in to it would show an overly inflated GC band on the application overview graph, because data from both web and non-web transactions would be included. This has been fixed, and GC time during web and non-web transactions is now tracked separately.

More accurate GC measurements on multi-threaded web servers

The agent could previously have reported inaccurate GC times on multi-threaded web servers such as Puma. It will now correctly report GC timings in multi-threaded contexts.

Improved ActiveMerchant instrumentation

The agent will now trace the store, unstore, and update methods on ActiveMerchant gateways. In addition, a bug preventing ActiveMerchant instrumentation from working on Ruby 1.9+ has been fixed.

Thanks to Troex Nevelin for the contribution!

More robust Real User Monitoring script injection with charset meta tags

Previous versions of the agent with Real User Monitoring enabled could have injected JavaScript code into the page above a charset meta tag. By the HTML5 spec, the charset tag must appear in the first 1024 bytes of the page, so the Ruby agent will now attempt to inject RUM script after a charset tag, if one is present.

More robust connection sequence with New Relic servers

A rare bug that could cause the agent's initial connection handshake with New Relic servers to silently fail has been fixed, and better logging has been added to the related code path to ease diagnosis of any future issues.

Prevent over-counting of queue time with nested transactions

When using add_transaction_tracer on methods called from within a Rails or Sinatra action, it was previously possible to get inflated queue time measurements, because queue time would be recorded for both the outer transaction (the Rails or Sinatra action) and the inner transaction (the method given to add_transaction_tracer). This has been fixed, so only the outermost transaction will now record queue time.