Sidekiq: perform_async and order-dependent operations - ruby-on-rails

There's a controller action in my Rails app that contacts a user via text-message and email. For reasons I won't go into, the text-message needs to complete before the email can be sent successfully. I originally had something like this:
controller:
class MyController < ApplicationController
def contact_user
ContactUserWorker.perform_async(#user.id)
end
end
workers:
class ContactUserWorker
include Sidekiq::Worker
def perform(user_id)
SendUserTextWorker.perform_async(user_id)
SendUserEmailWorker.perform_async(user_id)
end
end
class SendUserTextWorker
include Sidekiq::Worker
def perform(user_id)
user = User.find(user_id)
user.send_text
end
end
class SendUserEmailWorker
include Sidekiq::Worker
def perform(user_id)
user = User.find(user_id)
user.send_email
end
end
This was unreliable; sometimes the email would fail, sometimes both would fail. I'm trying to determine whether perform_async was the cause of the problem. Was the async part allowing the email to fire off before the text had completed? I'm a little fuzzy on how exactly perform_async works, but that sounded like a reasonable guess.
At first, I refactored ContactUserWorker to:
class ContactUserWorker
include Sidekiq::Worker
def perform(user_id)
user = User.find(user_id)
User.send_text
SendUserEmailWorker.perform_async(user_id)
end
end
Eventually though, I just moved the call to send_text out of the workers altogether and into the controller:
class MyController < ApplicationController
def contact_user
#user.send_text
SendUserEmailWorker.perform_async(#user.id)
end
end
This is a simplified version of the real code, but that's the gist of it. It seems to be working fine now, though I still wonder whether the problem was Sidekiq-related or if something else was going on.
I'm curious whether my original structure would've worked if I'd used perform instead of perform_async for all the calls except the email call. Like this:
class MyController < ApplicationController
def contact_user
ContactUserWorker.perform(#user.id)
end
end
class ContactUserWorker
include Sidekiq::Worker
def perform(user_id)
SendUserTextWorker.perform(user_id)
SendUserEmailWorker.perform_async(user_id)
end
end

If the email can only be sent after the text message has been sent, then send the email after successful completion of sending the text.
class ContactUserWorker
include Sidekiq::Worker
def perform(user_id)
SendUserTextWorker.perform_async(user_id)
end
end
class SendUserTextWorker
include Sidekiq::Worker
def perform(user_id)
user = User.find(user_id)
text_sent = user.send_text
SendUserEmailWorker.perform_async(user_id) if text_sent
end
end
class SendUserEmailWorker
include Sidekiq::Worker
def perform(user_id)
user = User.find(user_id)
user.send_email
end
end
In user.send_text you need to handle the fact that neither the text or the email has been sent.

I'm curious whether my original structure would've worked if I'd used perform instead of perform_async for all the calls except the email call
It would have. But this is not what you actually intdending. What you really want is this:
class MyController < ApplicationController
def contact_user
ContactUserWorker.perform_async(#user.id)
end
end
class ContactUserWorker
include Sidekiq::Worker
attr_reader :user_id
def perform(user_id)
#user_id = user_id
user.send_text
user.send_email
end
def user
#user ||= User.find user_id
end
end
The problem was indeed the perform async part. It schedules both tasks to be executed in the background by a separate sidekiq daemon process. i guess your sidekiq is configured to execute the jobs concurrently. In the first version you've first scheduled the ContactUserWorker to perform it's job in a background outside of the current rails request. As this worker is startet later on, it kicks off two separate delayed workers in turn, which are then run in parallel and so there is no way to determine which of the both executes/finishes first.
I don't know what you mean exatly by sending text, but sending an email is an io blocking process and therefore it was a good idea to perform this in a background, because it would be blocking a complete rails process otherwise until the email is delivered (on a typical unicorn/passenger multi-process deployment). And as you actually want to execute both tasks sequentially and as an atomic operation, it's totally fine, performing them by a single sidekiq job/worker.
You also don't have to check if send_text succeeds. Sidekiq will retry the complete job if any part of it fails with an exception

Related

delete request with sidekiq

I am new to Ruby on Rails and Sidekiq. I want to set this delete request to be done in Sidekiq queue and I don't know how to send it to the perform method, I am sending the Book model to the perform method
My controller Action code
def destroy
BaseWorkerJob.perform_async(Book)
end
My BaseWorkerJob class code
class BaseWorkerJob
include Sidekiq::Job
sidekiq_options retry:0
def perform(book)
# Do something
book.find(params[:id]).destroy!
sleep 15
end
end
SideKiq Error
enter image description here
ruby 3.1.2
Rails 7.0.4
You can send the model name and object id to the worker
def destroy
BaseWorkerJob.perform_async(Book.to_s, params[:id])
end
class BaseWorkerJob
include Sidekiq::Job
sidekiq_options retry: 0
def perform(klass_name, object_id)
klass_name.constantize.find(object_id).destroy!
end
end
Try it out!

Rails Service never called

I've written a simple service to create the data I want to send over my ActionCable connection, but judging from the development log it looks like it is never executed and I can't figure out why.
#app\channels\quiz_data_channel.rb:
class QuizDataChannel < ApplicationCable::Channel
def subscribed
stream_from specific_channel
end
def send_data
logger.debug "[AC] send data method entered" #<-- this part is executed
QuizDataCreation.new(user: current_user).create #<-- calling the service HERE
logger.debug "[AC] service created new quiz data" #<-- not executed
#after I fix this issue I would broadcast the data here
end
private
def specific_channel
"quiz_data_#{params[:access_key]}"
end
end
#app\services\quiz_data_creation.rb:
module Services
class QuizDataCreation
def initialize(user)
self.user = user
logger.debug "[AC] service - initialize user: #{self.user.inspect}" #<-- not executed
end
def create
logger.debug "[AC] Service entered!" #<-- not executed
end
end
end
Calling the send_data method works, and the first text ("[AC] send data method entered") is printed into the development log, but that's it. I've been looking at several tutorials and tried placing the service in a subfolder of models at first, and calling Services::QuizDataCreation.new(user: current_user).create , but haven't gotten it to work yet. I'm sure there is a pretty obvious solution, but I just can't see it so I would be really thankful for any pointers.
Try defining the service class without module names Services as below:
#app\services\quiz_data_creation.rb:
class QuizDataCreation
def initialize(user)
self.user = user
logger.debug "[AC] service - initialize user: #{self.user.inspect}"
end
def create
logger.debug "[AC] Service entered!"
end
end
And then, you can access it normally as:
QuizDataCreation.new(current_user).create

Accessing current_user in a model in a Rails 3.2 app

I have a Rails 3.2 app. It is a publishing app where we kick off several Sidekiq jobs in response to changes in content. I was calling this from the controller but there's now getting to be multiple points of entry and are now duplicating logic in multiple controllers. The proper place for this to be is in a callback in the model. However, accessing current_user is frowned upon in the model but for things like logging changes or app events, it is critical.
So I have two questions (1) Is there something I'm missing regarding the argument about accessing current_user when you want to be logging changes across complex model structures? and (2) Is the proposed solution here an effective one with last update over 2 years ago in terms of thread-safety? I use a three Unicorn processes on Heroku. https://stackoverflow.com/a/2513456/152825
Edit 1
Thinking through this, wondering if I should just do something like this in my application.rb
class ArcCurrentUser
#current_user_id
def self.id
return #current_user_id
end
def self.id=id_val
#current_user_id=id_val
end
end
and then in my current_user method in application_controller, just update ArcCurrentUser.id to #current_user.id? I will only be using it for this logging functionality.
You're correct in that you can't access current_user from a model.
As for the answer you linked, I'm not entirely sure but I think it's not fully thread-safe. From the same question, I like this answer https://stackoverflow.com/a/12713768/4035338 more.
Say we have a controller with this action
...
def update
#my_object = MyModel.find(params[:id])
#my_object.current_user = current_user
#my_object.assign_attributes params[:my_model]
#my_object.save
end
...
and this model
class MyModel < ActiveRecord::Base
attr_accessor :current_user
before_save :log_who_did_it
private
def log_who_did_it
return unless current_user.present?
puts "It was #{current_user}!"
end
end
Or my favourite
...
def update
#my_object = MyModel.find(params[:id])
#my_object.update_and_log_user(params[:my_model], current_user)
end
...
class MyModel < ActiveRecord::Base
...
def update_and_log_user(params, user)
update_attributes(params)
puts "It was #{user}!" if user.present?
end
end

Rails: Call sign_out(#user) from a background worker

I want to sign_out a given user from a background job? E.g. sign out a given user fom a Sidekiq background worker. Any idea how I can access the method sign_out from the worker?
class SessionWorker
include Sidekiq::Worker
sidekiq_options retry: false
def perform(user_id)
# user = User.find(user_id)
# sign_out(user) if user
end
end
end
Not 100% sure, but:
I don't think this is possible because when you run code from a background worker, that code is executed in a different context than your web application and thus has no access to the warden session where the login is 'registrated'.

rails 3: how to abort delivery method in actionmailer?

In my mailer controller, under certain conditions (missing data) we abort sending the email.
How do I exit the controller method without still rendering a view in that case?
return if #some_email_data.nil?
Doesn't do the trick since the view is still rendered (throwing an error every place I try to use #some_email_data unless I add a lot of nil checks)
And even if I do the nil checks, it complains there's no 'sender' (because I supposed did a 'return' before getting to the line where I set the sender and subject.
Neither does render ... return
Basically, RETURN DOESN'T RETURN inside a mailer method!
A much simpler solution than the accepted answer would be something like:
class SomeMailer < ActionMailer::Base
def some_method
if #some_email_data.nil?
self.message.perform_deliveries = false
else
mail(...)
end
end
end
If you're using Rails 3.2.9 (or later things even better) - there you can finally conditionally call mail(). Here's the related GitHub thread. Now the code can be reworked like this:
class SomeMailer < ActionMailer::Base
def some_method
unless #some_email_data.nil?
mail(...)
end
end
end
I just encountered same thing here.
My solution was following:
module BulletproofMailer
class BlackholeMailMessage < Mail::Message
def self.deliver
false
end
end
class AbortDeliveryError < StandardError
end
class Base < ActionMailer::Base
def abort_delivery
raise AbortDeliveryError
end
def process(*args)
begin
super *args
rescue AbortDeliveryError
self.message = BulletproofMailer::BlackholeMailMessage
end
end
end
end
Using these wrapper mailer would look like this:
class EventMailer < BulletproofMailer::Base
include Resque::Mailer
def event_created(event_id)
begin
#event = CalendarEvent.find(event_id)
rescue ActiveRecord::RecordNotFound
abort_delivery
end
end
end
It is also posted in my blog.
I've found this method that seems the least-invasive, as it works across all mailer methods without requiring you to remember to catch an error. In our case, we just want a setting to completely disable mailers for certain environments. Tested in Rails 6, although I'm sure it'll work just fine in Rails 5 as well, maybe lower.
class ApplicationMailer < ActionMailer::Base
class AbortDeliveryError < StandardError; end
before_action :ensure_notifications_enabled
rescue_from AbortDeliveryError, with: -> {}
def ensure_notifications_enabled
raise AbortDeliveryError.new unless <your_condition>
end
...
end
The empty lambda causes Rails 6 to just return an ActionMailer::Base::NullMail instance, which doesn't get delivered (same as if your mailer method didn't call mail, or returned prematurely).
Setting self.message.perform_deliveries = false did not work for me.
I used a similar approach as some of the other answers - using error handling to control the flow and prevent the mail from being sent.
The example below is aborting mail from being sent in non-Production ENVs to non-whitelisted emails, but the helper method logic can be whatever you need for your scenario.
class BaseMailer < ActionMailer::Base
class AbortedMailer < StandardError; end
def mail(**args)
whitelist_mail_delivery(args[:to])
super(args)
rescue AbortedMailer
Rails.logger.info "Mail aborted! We do not send emails to external email accounts outside of Production ENV"
end
private
def whitelist_mail_delivery(to_email)
return if Rails.env.production?
raise AbortedMailer.new unless internal_email?(to_email)
end
def internal_email?(to_email)
to_email.include?('#widgetbusiness.com')
end
end
I just clear the #to field and return, so deliver aborts when it doesn't have anything there. (Or just return before setting #to).
I haven't spent much time with rails 3 but you could try using
redirect_to some_other_route
alternatively, if you're really just checking for missing data you could do a js validation of the form fields and only submit if it passes.

Resources