Files
gitlab-foss/lib/gitlab/error_tracking.rb
2025-06-11 09:10:15 +00:00

231 lines
8.7 KiB
Ruby

# frozen_string_literal: true
module Gitlab
module ErrorTracking
# Exceptions in this group will receive custom Sentry fingerprinting
CUSTOM_FINGERPRINTING = %w[
Acme::Client::Error::BadNonce
Acme::Client::Error::NotFound
Acme::Client::Error::RateLimited
Acme::Client::Error::Timeout
Acme::Client::Error::UnsupportedOperation
ActiveRecord::ConnectionTimeoutError
Gitlab::RequestContext::RequestDeadlineExceeded
GRPC::DeadlineExceeded
JIRA::HTTPError
Rack::Timeout::RequestTimeoutException
].freeze
PROCESSORS = [
::Gitlab::ErrorTracking::Processor::SidekiqProcessor,
::Gitlab::ErrorTracking::Processor::GrpcErrorProcessor,
::Gitlab::ErrorTracking::Processor::ContextPayloadProcessor,
::Gitlab::ErrorTracking::Processor::SanitizeErrorMessageProcessor,
# IMPORTANT: this processor must stay at the bottom, right before
# sending the event to Sentry.
::Gitlab::ErrorTracking::Processor::SanitizerProcessor
].freeze
class << self
def configure
Sentry.init do |config|
config.dsn = sentry_dsn
config.release = Gitlab.revision
config.environment = sentry_environment
config.before_send = method(:before_send)
config.background_worker_threads = 0
config.send_default_pii = true
config.send_modules = false
config.traces_sample_rate = 0.2 if Gitlab::Utils.to_boolean(ENV['ENABLE_SENTRY_PERFORMANCE_MONITORING'])
# Reason for disabling the below configs https://gitlab.com/gitlab-org/gitlab/-/merge_requests/150771#note_1881953691
config.metrics.enabled = false
config.metrics.enable_code_locations = false
config.propagate_traces = false
config.trace_propagation_targets = []
config.enabled_patches = [:sidekiq_cron]
config.enable_tracing = false
yield config if block_given?
end
end
# This should be used when you want to passthrough exception handling:
# rescue and raise to be caught in upper layers of the application.
#
# If the exception implements the method `sentry_extra_data` and that method
# returns a Hash, then the return value of that method will be merged into
# `extra`. Exceptions can use this mechanism to provide structured data
# to sentry in addition to their message and back-trace.
def track_and_raise_exception(exception, extra = {}, tags = {})
process_exception(exception, extra: extra, tags: tags)
raise exception
end
# This can be used for investigating exceptions that can be recovered from in
# code. The exception will still be raised in development and test
# environments.
#
# That way we can track down these exceptions with as much information as we
# need to resolve them.
#
# If the exception implements the method `sentry_extra_data` and that method
# returns a Hash, then the return value of that method will be merged into
# `extra`. Exceptions can use this mechanism to provide structured data
# to sentry in addition to their message and back-trace.
#
# Provide an issue URL for follow up.
# as `issue_url: 'http://gitlab.com/gitlab-org/gitlab/issues/111'`
def track_and_raise_for_dev_exception(exception, extra = {}, tags = {})
process_exception(exception, extra: extra, tags: tags)
raise exception if should_raise_for_dev?
end
# This should be used when you want to track the exception and not raise
# with the default trackers (Sentry and Logger).
#
# If the exception implements the method `sentry_extra_data` and that method
# returns a Hash, then the return value of that method will be merged into
# `extra`. Exceptions can use this mechanism to provide structured data
# to sentry in addition to their message and back-trace.
def track_exception(exception, extra = {}, tags = {})
process_exception(exception, extra: extra, tags: tags)
end
# This should be used when you only want to log the exception,
# but not send it to Sentry or raise.
#
# If the exception implements the method `sentry_extra_data` and that method
# returns a Hash, then the return value of that method will be merged into
# `extra`. Exceptions can use this mechanism to provide structured data
# to sentry in addition to their message and back-trace.
def log_exception(exception, extra = {})
process_exception(exception, extra: extra, trackers: [Logger])
end
# This should be used when you want to log the exception and passthrough
# exception handling: rescue and raise to be catched in upper layers of
# the application.
#
# If the exception implements the method `sentry_extra_data` and that method
# returns a Hash, then the return value of that method will be merged into
# `extra`. Exceptions can use this mechanism to provide structured data
# to sentry in addition to their message and back-trace.
def log_and_raise_exception(exception, extra = {})
process_exception(exception, extra: extra, trackers: [Logger])
raise exception
end
private
def before_send(event, hint)
# Don't report Sidekiq retry errors to Sentry
return if hint[:exception].is_a?(Gitlab::SidekiqMiddleware::RetryError)
inject_context_for_exception(event, hint[:exception])
custom_fingerprinting(event, hint[:exception])
PROCESSORS.reduce(event) do |processed_event, processor|
processor.call(processed_event)
end
end
def process_exception(exception, extra:, tags: {}, trackers: default_trackers)
Gitlab::Utils.allow_within_concurrent_ruby do
context_payload = Gitlab::ErrorTracking::ContextPayloadGenerator.generate(exception, extra, tags)
trackers.each do |tracker|
tracker.capture_exception(exception, **context_payload)
end
end
end
def default_trackers
[].tap do |destinations|
# There is a possibility that this method is called before Sentry is
# configured. Since Sentry 4.0, some methods of Sentry are forwarded to
# to `nil`, hence we have to check the client as well.
destinations << ::Sentry if ::Sentry.get_current_client && ::Sentry.configuration.dsn
destinations << Logger
end
end
# Some configuration attributes like `dsn`, and `environment`
# can be configured both via `ENV` and `Application Settings`.
# The reason being is while GitLab.com uses application_settings
# in Geo installations, we can't override values in the primary database.
# Setting this value in application_settings would propagate the value
# to all Geo nodes, which doesn't solve that particular problem.
def sentry_dsn
env_sentry_dsn || database_sentry_dsn
end
def sentry_environment
env_sentry_environment || database_sentry_environment
end
def database_sentry_dsn
return unless sentry_configurable?
return unless database_sentry_enabled?
Gitlab::CurrentSettings.sentry_dsn
end
def env_sentry_dsn
return unless sentry_configurable?
return unless env_sentry_enabled?
Gitlab.config.sentry.dsn
end
def env_sentry_environment
return unless sentry_configurable?
return unless env_sentry_enabled?
Gitlab.config.sentry.environment
end
def database_sentry_environment
return unless sentry_configurable?
return unless database_sentry_enabled?
return unless Gitlab::CurrentSettings.respond_to?(:sentry_environment)
Gitlab::CurrentSettings.sentry_environment
end
def database_sentry_enabled?
Gitlab::CurrentSettings.respond_to?(:sentry_enabled?) &&
Gitlab::CurrentSettings.sentry_enabled?
end
def env_sentry_enabled?
Gitlab.config.sentry.enabled
end
def sentry_configurable?
Rails.env.production? || Rails.env.development?
end
def should_raise_for_dev?
Rails.env.development? || Rails.env.test?
end
# Group common, mostly non-actionable exceptions by type and message,
# rather than cause
def custom_fingerprinting(event, ex)
return event unless CUSTOM_FINGERPRINTING.include?(ex.class.name)
event.fingerprint = [ex.class.name, ex.message]
end
def inject_context_for_exception(event, ex)
sql = Gitlab::ExceptionLogFormatter.find_sql(ex)
event.extra[:sql] = sql if sql
end
end
end
end