5.17.3

Internal

• Update key, unit and tags sanitization logic for metrics #2292
• Consolidate client report and rate limit handling with data categories #2294
• Record :network_error client reports for send_envelope #2295

Bug Fixes

• Make sure isolated envelopes respect config.enabled_environments #2291
  • Fixes #2287

5.17.2

Internal

• Add Mechanism interface and default to unhandled for integration exceptions #2280

Bug Fixes

• Don't instantiate connection in ActiveRecordSubscriber (#2278)

5.17.1

Bug Fixes

• Fix NoMethodError / Make session_tracking check consistent (#2269)

5.17.0

Features

• Add support for distributed tracing in sentry-delayed_job #2233
• Fix warning about default gems on Ruby 3.3.0 (#2225)
• Add hint: support to Sentry::Rails::ErrorSubscriber #2235
• Add Metrics support

  • Add main APIs and Aggregator thread #2247

  • Add Sentry::Metrics.timing API for measuring block duration #2254

  • Add metric summaries on spans #2255

  • Add config.metrics.before_emit callback #2258

  • Add code locations for metrics #2263

    The SDK now supports recording and aggregating metrics. A new thread will be started
    for aggregation and will flush the pending data to Sentry every 5 seconds.

    To enable this behavior, use:

    Sentry.init do |config|
      # ...
      config.metrics.enabled = true
    end

    And then in your application code, collect metrics as follows:

    # increment a simple counter
    Sentry::Metrics.increment('button_click')
    # set a value, unit and tags
    Sentry::Metrics.increment('time', 5, unit: 'second', tags: { browser:' firefox' })
    
    # distribution - get statistical aggregates from an array of observations
    Sentry::Metrics.distribution('page_load', 15.0, unit: 'millisecond')
    
    # gauge - record statistical aggregates directly on the SDK, more space efficient
    Sentry::Metrics.gauge('page_load', 15.0, unit: 'millisecond')
    
    # set - get unique counts of elements
    Sentry::Metrics.set('user_view', 'jane')
    
    # timing - measure duration of code block, defaults to seconds
    # will also automatically create a `metric.timing` span
    Sentry::Metrics.timing('how_long') { sleep(1) }
    # timing - measure duration of code block in other duraton units
    Sentry::Metrics.timing('how_long_ms', unit: 'millisecond') { sleep(0.5) }

    You can filter some keys or update tags on the fly with the before_emit callback, which will be triggered before a metric is aggregated.

    Sentry.init do |config|
      # ...
      # the 'foo' metric will be filtered and the tags will be updated to add :bar and remove :baz
      config.metrics.before_emit = lambda do |key, tags|
        return nil if key == 'foo'
        tags[:bar] = 42
        tags.delete(:baz)
        true
      end
    end

    By default, the SDK will send code locations for unique metrics (defined by type, key and unit) once a day and with every startup/shutdown of your application.
    You can turn this off with the following:

    Sentry.init do |config|
      # ...
      config.metrics.enable_code_locations = false
    end

Bug Fixes

• Fix undefined method 'constantize' issue in sentry-resque (#2248)
• Only instantiate SessionFlusher when the SDK is enabled under the current env #2245
  • Fixes #2234
• Update backtrace parsing regexp to support Ruby 3.4 (#2252)
• Make sure sending_allowed? is respected irrespective of spotlight configuration (#2231)
  • Fixes #2226

5.16.1

Bug Fixes

• Pin sqlite3 gem for building because of failed release #2222

5.16.0

Features

• Add backpressure handling for transactions #2185

  The SDK can now dynamically downsample transactions to reduce backpressure in high
  throughput systems. It starts a new BackpressureMonitor thread to perform some health checks
  which decide to downsample (halved each time) in 10 second intervals till the system
  is healthy again.

  To enable this behavior, use:

  Sentry.init do |config|
    # ...
    config.traces_sample_rate = 1.0
    config.enable_backpressure_handling = true
  end

  If your system serves heavy load, please let us know how this feature works for you!

• Implement proper flushing logic on close for Client Reports and Sessions #2206

• Support cron with timezone for sidekiq-scheduler patch #2209

  • Fixes #2187

• Add Cron::Configuration object that holds defaults for all MonitorConfig objects #2211

  Sentry.init do |config|
    # ...
    config.cron.default_checkin_margin = 1
    config.cron.default_max_runtime = 30
    config.cron.default_timezone = 'America/New_York'
  end

• Clean up logging #2216

• Pick up config.cron.default_timezone from Rails config #2213

• Don't add most scope data (tags/extra/breadcrumbs) to CheckInEvent #2217

5.15.2

Bug Fixes

• Fix sample_rate applying to check-in events #2203
  • Fixes #2202

5.15.1

Features

• Expose configuration.background_worker_max_queue to control thread pool queue size #2195

Bug Fixes

• Fix Sentry::Cron::MonitorCheckIns monkeypatch keyword arguments #2199
  • Fixes #2198

5.15.0

Features

• You can now use Spotlight with your apps that use sentry-ruby! #2175
• Improve default slug generation for sidekiq-scheduler #2184

Bug Fixes

• Network errors raised in Sentry::HTTPTransport will no longer be reported to Sentry #2178

5.14.0

Features

• Improve default slug generation for Crons #2168
• Change release name generator to use full SHA commit hash and align with sentry-cli and other Sentry SDKs #2174
• Automatic Crons support for scheduling gems

  • Add support for sidekiq-cron #2170

    You can opt in to the sidekiq-cron patch and we will automatically monitor check-ins for all jobs listed in your config/schedule.yml file.

    config.enabled_patches += [:sidekiq_cron]

  • Add support for sidekiq-scheduler #2172

    You can opt in to the sidekiq-scheduler patch and we will automatically monitor check-ins for all repeating jobs (i.e. cron, every, and interval) specified in the config.

    config.enabled_patches += [:sidekiq_scheduler]

Bug Fixes

• Fixed a deprecation in sidekiq-ruby error handler #2160
• Avoid invoking ActiveSupport::BroadcastLogger if not defined #2169
• Respect custom Delayed::Job.max_attempts if it's defined #2176
• Fixed a bug where Net::HTTP instrumentation won't work for some IPv6 addresses #2180
• Allow non-string error message to be reported to sentry (#2137)