Sentry Ruby/ Rails not capturing exceptions from rails/runner

[ckelly]

Describe the bug

exceptions in rails/runner jobs don't appear to be getting captured

To Reproduce
from project command line

./bin/rails runner 'Sentry.capture_message("test sentry runner capture")'

swapping back to sentry-raven (= 3.1.1) with below raven config and attempting:

./bin/rails runner 'Raven.capture_message("test old raven runner capture")'

sends messages to our sentry dashboard as expected.

(I've added development to environments in order to test locally, of course).

I've also verified that the jobs are getting the proper environment at runtime to match one of the expected/ enabled envs.

potential regression from raven fix - #325

Expected behavior
Exceptions/ messages log to our sentry project

Actual behavior
No exceptions or messages are being logged

Environment

• Ruby Version: 2.7.1
• SDK Version:
• sentry-rails (= 4.2.2)
• sentry-ruby (= 4.2.2)
• sentry-sidekiq (= 4.2.1)
• Rails 6.0.3.5
• Sidekiq 6.1.3
• Whenever 1.0.0 (cron job creation)

Current Sentry Config

This is not necessary but could be helpful.

Sentry.init do |config|
  config.dsn = 'MY_DSN'
  config.enabled_environments = %w[production staging]
  config.environment = "#{Rails.env}-console" if defined?(Rails::Console)
  config.breadcrumbs_logger = %i[sentry_logger active_support_logger]

  # set a uniform sample rate between 0.0 and 1.0
  config.traces_sample_rate = 0.2

end

Old Raven Config

Raven.configure do |config|
  config.dsn = 'MY_DSN'
  config.sanitize_fields = Rails.application.config.filter_parameters.map(&:to_s)
  config.environments = %w[production staging]

  config.current_environment = "#{Rails.env}-console" if defined?(Rails::Console)
end

[st0012]

this is because sentry-ruby now sends events via the background worker by default (powered by a thread pool). and it's likely that the runner process exists before the worker sends out the event. you can work around this in 2 ways:

1. use hint: { background: false } to force it send events synchronously. like Sentry.capture_message("test sentry runner capture", hint: { background: false }).
2. set config.background_worker_threads = 0 if you want to disable the background worker all at once.

I'll find a way to avoid such problems. but for now, there doesn't seem to be a way for solving it cleanly.

[ckelly]

thanks for the response. I think we'll capture the errors in our job base class and send them synchronously for now, and keep the rest of our stack async.

[ckelly]

As a followup, we implemented the catch and sync-send in our runner job base class, and test exception reporting is much more consistent.

[st0012]

this is because sentry-ruby now sends events via the background worker by default (powered by a thread pool). and it's likely that the runner process exists before the worker sends out the event. you can work around this in 2 ways:

1. use hint: { background: false } to force it send events synchronously. like Sentry.capture_message("test sentry runner capture", hint: { background: false }).
2. set config.background_worker_threads = 0 if you want to disable the background worker all at once.

I'll find a way to avoid such problems. but for now, there doesn't seem to be a way for solving it cleanly.

[ckelly]

thanks for the response. I think we'll capture the errors in our job base class and send them synchronously for now, and keep the rest of our stack async.

[ckelly]

As a followup, we implemented the catch and sync-send in our runner job base class, and test exception reporting is much more consistent.