discourse/config/initializers/100-logster.rb
David Taylor 374ab82dd6
FIX: Ensure 100-logster initializer is run before 101-lograge (#12455)
The logster initializer tries to adds RailsMultisite::Formatter to the STDOUT logger. In production, the lograge initializer then removes the RailsMultisite:Formatter because the JSON log will include the database.

e10a74694a used `Rails.application.reloader.to_prepare` to defer running the 100-logster initializer, which meant it ran **after** 101-lograge. This meant that we were writing JSON logs with a non-json text prefix.

The `to_prepare` was added because our freedom-patches are now deferred using `to_prepare`, and some initializers were relying on the freedom patches. However, following 1533cbb38b, we decided to load the RailsMultisite freedom patch without `to_prepare`. Therefore, `005-site_settings` and `100-logster` no longer need to use `to_prepare`. Removing it means that these initializers are back to running in sequential order, and the logging issue will be resolved.

The only remaining initializer which depends on freedom patches is `100-i18n`. I've added a comment to explain why.
2021-03-19 19:51:13 +00:00

151 lines
5.2 KiB
Ruby

# frozen_string_literal: true
if GlobalSetting.skip_redis?
Rails.application.reloader.to_prepare do
if Rails.logger.respond_to? :chained
Rails.logger = Rails.logger.chained.first
end
end
return
end
if Rails.env.development? && RUBY_VERSION.match?(/^2\.5\.[23]/)
STDERR.puts "WARNING: Discourse development environment runs slower on Ruby 2.5.3 or below"
STDERR.puts "We recommend you upgrade to Ruby 2.6.1 for the optimal development performance"
# we have to used to older and slower version of the logger cause the new one exposes a Ruby bug in
# the Queue class which causes segmentation faults
Logster::Scheduler.disable
end
if Rails.env.development? && !Sidekiq.server? && ENV["RAILS_LOGS_STDOUT"] == "1"
console = ActiveSupport::Logger.new(STDOUT)
original_logger = Rails.logger.chained.first
console.formatter = original_logger.formatter
console.level = original_logger.level
unless ActiveSupport::Logger.logger_outputs_to?(original_logger, STDOUT)
original_logger.extend(ActiveSupport::Logger.broadcast(console))
end
end
if Rails.env.production?
Logster.store.ignore = [
# honestly, Rails should not be logging this, its real noisy
/^ActionController::RoutingError \(No route matches/,
/^PG::Error: ERROR:\s+duplicate key/,
/^ActionController::UnknownFormat/,
/^ActionController::UnknownHttpMethod/,
/^AbstractController::ActionNotFound/,
# ignore any empty JS errors that contain blanks or zeros for line and column fields
#
# Line:
# Column:
#
/(?m).*?Line: (?:\D|0).*?Column: (?:\D|0)/,
# suppress empty JS errors (covers MSIE 9, etc)
/^(Syntax|Script) error.*Line: (0|1)\b/m,
# CSRF errors are not providing enough data
# suppress unconditionally for now
/^Can't verify CSRF token authenticity.$/,
# Yandex bot triggers this JS error a lot
/^Uncaught ReferenceError: I18n is not defined/,
# related to browser plugins somehow, we don't care
/Error calling method on NPObject/,
# 404s can be dealt with elsewhere
/^ActiveRecord::RecordNotFound/,
# bad asset requested, no need to log
/^ActionController::BadRequest/,
# we can't do anything about invalid parameters
/Rack::QueryParser::InvalidParameterError/,
# we handle this cleanly in the message bus middleware
# no point logging to logster
/RateLimiter::LimitExceeded.*/m,
# see https://github.com/rails/rails/issues/34599
# Poll defines an enum with the value `open` ActiveRecord then attempts
# AR then warns cause #open is being redefined, it is already defined
# privately in Kernel per: http://ruby-doc.org/core-2.5.3/Kernel.html#method-i-open
# Once the rails issue is fixed we can stop this error suppression and stop defining
# scopes for the enums
/^Creating scope :open\. Overwriting existing method Poll\.open\./,
]
Logster.config.env_expandable_keys.push(:hostname, :problem_db)
end
Logster.store.max_backlog = GlobalSetting.max_logster_logs
# TODO logster should be able to do this automatically
Logster.config.subdirectory = "#{GlobalSetting.relative_url_root}/logs"
Logster.config.application_version = Discourse.git_version
Logster.config.enable_custom_patterns_via_ui = true
Logster.config.enable_js_error_reporting = GlobalSetting.enable_js_error_reporting
store = Logster.store
redis = Logster.store.redis
store.redis_prefix = Proc.new { redis.namespace }
store.redis_raw_connection = redis.without_namespace
severities = [Logger::WARN, Logger::ERROR, Logger::FATAL, Logger::UNKNOWN]
RailsMultisite::ConnectionManagement.each_connection do
error_rate_per_minute = SiteSetting.alert_admins_if_errors_per_minute rescue 0
if (error_rate_per_minute || 0) > 0
store.register_rate_limit_per_minute(severities, error_rate_per_minute) do |rate|
MessageBus.publish("/logs_error_rate_exceeded",
{
rate: rate,
duration: 'minute',
publish_at: Time.current.to_i
},
group_ids: [Group::AUTO_GROUPS[:admins]]
)
end
end
error_rate_per_hour = SiteSetting.alert_admins_if_errors_per_hour rescue 0
if (error_rate_per_hour || 0) > 0
store.register_rate_limit_per_hour(severities, error_rate_per_hour) do |rate|
MessageBus.publish("/logs_error_rate_exceeded",
{
rate: rate,
duration: 'hour',
publish_at: Time.current.to_i,
},
group_ids: [Group::AUTO_GROUPS[:admins]]
)
end
end
end
if Rails.configuration.multisite
if Rails.logger.respond_to? :chained
chained = Rails.logger.chained
chained && chained.first.formatter = RailsMultisite::Formatter.new
end
end
Logster.config.project_directories = [
{ path: Rails.root.to_s, url: "https://github.com/discourse/discourse", main_app: true }
]
Discourse.plugins.each do |plugin|
next if !plugin.metadata.url
Logster.config.project_directories << {
path: "#{Rails.root.to_s}/plugins/#{plugin.directory_name}",
url: plugin.metadata.url
}
end