Rails on Google Cloud Run
- Logging
- Error Reporting
- Active Job Adapter via Cloud Tasks
- Minor patches for better compatibility
- Works with Ruby 3 and Rails 6
Usage
logger.info "Hello World"
logger.info do
"Expensive logging operation, only run when logged"
end
logger.info "Labels work, too!", my_label: "works", another_one: "great"
All Google Cloud Logging Severities are supported:
logger.default (or logger.unknown) - The log entry has no assigned severity level.
logger.debug - Debug or trace information.
logger.info - Routine information, such as ongoing status or performance.
logger.notice - Normal but significant events, such as start up, shut down, or a configuration change.
logger.warning (or logger.warn) - Warning events might cause problems.
logger.error - Error events are likely to cause problems.
logger.critical (or logger.fatal) - Critical events cause more severe problems or outages.
logger.alert - A person must take an action immediately.
logger.emergency - One or more systems are unusable.
Installation
Add the gem to your Gemfile and run bundle install
.
# Gemfile
group :production do
gem 'google_cloud_run'
end
In your production config:
# config/environments/production.rb
config.log_level = :g_notice
config.logger = GoogleCloudRun::Logger.new
config.active_job.queue_adapter = :google_cloudrun_tasks
config.google_cloudrun.job_queue_default_region = "us-central1"
config.google_cloudrun.job_callback_url = "https://your-domain.com/rails/google_cloudrun/job_callback"
Set the default queue:
# app/jobs/application_job.rb
queue_as "my-queue"
# or if `config.google_cloudrun.job_queue_default_region` isn't set:
queue_as "us-central1/my-queue"
# config/environments/production.rb
config.action_mailer.deliver_later_queue_name = "my-queue" # or "us-central1/my-queue"
In the default production config, the logger is wrapped around
a ENV["RAILS_LOG_TO_STDOUT"].present?
block. I usually just
remove this block so I don't have to actually set this ENV var.
You can also remove config.log_formatter
and config.log_tags
as we don't need it anymore.
I recommend logging :g_notice
and higher. Rails logs a lot of noise when logging
:info
and higher.
Configuration
You can change more settings in config/environments/production.rb
. See below
for the default configuration.
# Enable Google Cloud Logging
config.google_cloudrun.logger = true
# Set output (STDERR or STDOUT)
config.google_cloudrun.out = STDERR
# Add source location (file, line number, method) to each log
config.google_cloudrun.logger_source_location = true
# Run Proc to assign current user as label to each log
config.google_cloudrun.logger_user = nil
# Enable Error Reporting
config.google_cloudrun.error_reporting = true
# Assign a default severity level to exceptions
config.google_cloudrun.error_reporting_exception_severity = :critical
# Run Proc to assign current user to Error Report
config.google_cloudrun.error_reporting_user = nil
# Turn logs into error reports for this severity and higher.
# Set to nil to disable.
config.google_cloudrun.error_reporting_level = :error
# When log is turned into error report, discard the original
# log and only report the error.
# Set to false to log and report the error at the same time.
config.google_cloudrun.error_reporting_discard_log = true
# Don't log or error report the following exceptions,
# because Cloud Run will create access logs for us already.
config.google_cloudrun.silence_exceptions = [
ActionController::RoutingError,
ActionController::MethodNotAllowed,
ActionController::UnknownHttpMethod,
ActionController::NotImplemented,
ActionController::UnknownFormat,
ActionController::BadRequest,
ActionController::ParameterMissing,
]
# Set Rails' request id to the trace id from X-Cloud-Trace-Context header
# as set by Cloud Run.
config.google_cloudrun.patch_request_id = true
# Enable Jobs via Cloud Tasks
config.google_cloudrun.jobs = true
# Set the default Google Cloud Task region, i.e. us-central1
config.google_cloudrun.job_queue_default_region = nil
# Google Cloud Tasks will call this url to execute the job
config.google_cloudrun.job_callback_url = nil # required, see above
# The default route for the callback url.
config.google_cloudrun.job_callback_path = "/rails/google_cloudrun/job_callback"
# Time for a job to run in seconds, default is 30min.
# Use `timeout_after 5.minutes` to configure a job individually.
config.google_cloudrun.job_timeout_sec = 1800 # (min 15s, max 30m)
Both error_reporting_user
and logger_user
expect a Proc like this:
config.google_cloudrun.logger_user = Proc.new do |request|
# extract and return user id from request, example:
request.try { cookie_jar.encrypted[:user_id] }
end
An example job:
class MyJob < ApplicationJob
queue_as "us-central1/urgent"
timeout_after 1.minute # min 15s, max 30m, overrides config.google_cloudrun.job_timeout_sec
def perform(*args)
# Do something
end
end
Cloud Task considerations
- Cloud Tasks are a better fit than Google Pub/Sub. Read more
- I'd recommend to create two different Cloud Run services.
One for HTTP requests (aka Heroku Dynos) and another service
for jobs (aka Heroku Workers). Set the
Request Timeout
for the request-bound service to something like15s
, and for workers to1800s
or matchconfig.google_cloudrun.job_timeout_sec
. - Cloud Task execution calls are authenticated with a Google-issued
OIDC token. So even though
/rails/google_cloudrun/job_callback
is publicly available, without a valid token, no job will be executed. - Cloud Task job processing is async. It supports multiple queues. Delayed jobs are natively supported through Cloud Task. Priority jobs are not supported, use different queues for that, i.e. "urgent", or "low-priority". Timeouts can be set globally or per job-basis (min 15s, max 30m). Retries are natively supported by Cloud Tasks.