Mission Control — Jobs

This gem provides a Rails-based frontend to Active Job adapters. It currently supports Resque and Solid Queue. Its features depend on those offered by the adapter itself. At a minimum, it allows you to inspect job queues and jobs currently waiting in those queues and inspect and retry or discard failed jobs.

Installation

Add this line to your application's Gemfile:

gem "mission_control-jobs"

And then execute:

$ bundle install

Basic configuration

Mount Mission Control Job's engine where you wish to have it accessible from your app, in your routes.rb file:

Rails.application.routes.draw do
  # ...
  mount MissionControl::Jobs::Engine, at: "/jobs"

And that's it. With this alone, you should be able to access Mission Control Job's UI, where you can browse the existing queues, jobs pending in these queues, jobs in different statuses, and discard and retry failed jobs:

Queues tab in a simple app

Failed jobs tab in a simple app

API-only apps or apps using vite_rails and other asset pipelines outside Rails

If you want to use this gem with an API-only Rails app or an app that's using vite_ruby/vite_rails, or some other custom asset pipeline different from Sprockets and Propshaft, you need one more thing: configure an asset pipeline so you can serve the JavaScript and CSS included in this gem. We recommend to use Propshaft. You simply need to add this line to your application's Gemfile:

gem "propshaft"

Then execute

$ bundle install

Then, make sure you add a step to your deployment pipeline to precompile assets:

RAILS_ENV=production rails assets:precompile

For example, if you're using the Dockerfile generated by Rails with an API-only app or having skipped the assets pipeline, re-add:

# Precompiling assets for production without requiring secret RAILS_MASTER_KEY
RUN SECRET_KEY_BASE_DUMMY=1 ./bin/rails assets:precompile

Note: Legacy CSS bundlers sass-rails and sassc-rails may fail to compile some of the CSS vendored into this library from Bulma, which was created in Dart SASS. You will therefore need to upgrade to dartsass-rails or some library that relies on it, like cssbundling-rails.

Authentication

Mission Control comes with HTTP basic authentication enabled and closed by default. Credentials are stored in Rails's credentials like this:

mission_control:
  http_basic_auth_user: dev
  http_basic_auth_password: secret

If no credentials are configured, Mission Control won't be accessible. To set these up, you can run the generator provided like this:

bin/rails mission_control:jobs:authentication:configure

To set them up for different environments you can use the RAILS_ENV environment variable, like this:

RAILS_ENV=production bin/rails mission_control:jobs:authentication:configure

Custom authentication

You can provide your own authentication mechanism, for example, if you have a certain type of admin user in your app that can access Mission Control. To make this easier, you can specify a different controller as the base class for Mission Control's controllers. By default, Mission Control's controllers will extend the host app's ApplicationController, but you can change this easily:

Rails.application.configure do
  MissionControl::Jobs.base_controller_class = "AdminController"
end

Or, in your environment config or application.rb:

config.mission_control.jobs.base_controller_class = "AdminController"

If you do this, you can disable the default HTTP Basic Authentication using the following option:

config.mission_control.jobs.http_basic_auth_enabled = false

Other configuration settings

Besides base_controller_class, you can also set the following for MissionControl::Jobs or config.mission_control.jobs:

  • logger: the logger you want Mission Control Jobs to use. Defaults to ActiveSupport::Logger.new(nil) (no logging). Notice that this is different from Active Job's logger or Active Job's backend's configured logger.
  • delay_between_bulk_operation_batches: how long to wait between batches when performing bulk operations, such as discard all or retry all jobs—defaults to 0
  • adapters: a list of adapters that you want Mission Control to use and extend. By default this will be the adapter you have set for active_job.queue_adapter.
  • internal_query_count_limit: in count queries, the maximum number of records that will be counted if the adapter needs to limit these queries. True counts above this number will be returned as INFINITY. This keeps count queries fast—defaults to 500,000
  • scheduled_job_delay_threshold: the time duration before a scheduled job is considered delayed. Defaults to 1.minute (a job is considered delayed if it hasn't transitioned from the scheduled status 1 minute after the scheduled time).
  • show_console_help: whether to show the console help. If you don't want the console help message, set this to false—defaults to true.
  • backtrace_cleaner: a backtrace cleaner used for optionally filtering backtraces on the Failed Jobs detail page. Defaults to Rails::BacktraceCleaner.new. See the Advanced configuration section for how to configure/override this setting on a per application/server basis.

This library extends Active Job with a querying interface and the following setting:

  • config.active_job.default_page_size: the internal batch size that Active Job will use when sending queries to the underlying adapter and the batch size for the bulk operations defined above—defaults to 1000.

Adapter Specifics

  • Resque: Queue pausing is supported only if you have resque-pause installed in your project
  • Solid Queue: Requires version >= 1.0.1.

Advanced configuration

When we built Mission Control Jobs, we did it with the idea of managing multiple apps' job backends from a single, centralized app that we used for monitoring, alerts and other tools that related to all our apps. Some of our apps run in more than one datacenter, and we run different Resque instances with different Redis configurations in each. Because of this, we added support for multiple apps and multiple adapters per app. Even when running Mission Control Job within the app it manages, and a single DC, as we migrated from Resque to Solid Queue, we needed to manage both adapters from Mission Control.

Without adding any additional configuration to the one described before, Mission Control will be configured with one single app and a single server for your configured active_job.queue_adapter.

If you want to support multiple adapters, you need to add them to Mission Control configuration via the adapters setting mentioned above. For example:

config.mission_control.jobs.adapters = [ :resque, :solid_queue ]

Then, to configure the different apps and/or different servers, you can do so in an initializer like this (taken from our dummy app for testing purposes):

require "resque"
require "resque_pause_helper"

require "solid_queue"

Resque.redis = Redis::Namespace.new "#{Rails.env}", redis: Redis.new(host: "localhost", port: 6379)

SERVERS_BY_APP = {
  BC4: %w[ resque_ashburn resque_chicago ],
  HEY: %w[ resque solid_queue ]
}

def redis_connection_for(app, server)
  redis_namespace = Redis::Namespace.new "#{app}:#{server}", redis: Resque.redis.instance_variable_get("@redis")
  Resque::DataStore.new redis_namespace
end

SERVERS_BY_APP.each do |app, servers|
  queue_adapters_by_name = servers.collect do |server|
    queue_adapter = if server.start_with?("resque")
      ActiveJob::QueueAdapters::ResqueAdapter.new(redis_connection_for(app, server))
    else
      ActiveJob::QueueAdapters::SolidQueueAdapter.new
    end

    # Default:
    # 
    # @return Array<String, ActiveJob::QueueAdapters::Base)
    # An array where:
    # * the String represents the symbolic name for this server within the UI
    # * ActiveJob::QueueAdapters::Base adapter instance used to access this Application Server/Service
    [ server, queue_adapter ] 

    # Optional return formats:
    # 
    # @return Array<String, Array<ActiveJob::QueueAdapters::Base>>
    # * This is equivalent, and behaves identically to, the format the default format above. 
    # [ server, [ queue_adapter ]]  # without optional backtrace cleaner
    # 
    # @return Array<String, Array<ActiveJob::QueueAdapters::Base, ActiveSupport::BacktraceCleaner>>
    # * This format adds an optional ActiveSupport::BacktraceCleaner to override the system wide
    #   backtrace cleaner for *this* Application Server/Service.
    # [ server, [ queue_adapter, BacktraceCleaner.new ]]  # with optional backtrace cleaner
  end.to_h

  MissionControl::Jobs.applications.add(app, queue_adapters_by_name)
end

This is an example for two different apps, BC4 and HEY, each one with two servers. BC4 has two Resque servers with two different configurations, and HEY has one Resque server and one Solid Queue server.

Currently, only one Solid Queue configuration is supported, but support for several Solid Queue backends (with different databases) is planned.

This is how we set Resque and Solid Queue together when we migrated from one to the other:

queue_adapters_by_name = {
  resque: ActiveJob::QueueAdapters.lookup(:resque).new, # This will use Resque.redis as the redis client
  solid_queue: ActiveJob::QueueAdapters.lookup(:solid_queue).new
}

MissionControl::Jobs.applications.add("hey", queue_adapters_by_name)

When you have multiple apps and servers configured, you can choose between them with select and toggle menus:

Queues tab with multiple apps and servers

Basic UI usage

As mentioned, the features available in Mission Control depend on the adapter you're using, as each adapter supports different features. Besides inspecting the queues and the jobs in them, and discarding and retrying failed jobs, you can inspect jobs in different statuses supported by each adapter, filter them by queue name and job class name (with the idea of adding more filters in the future), pause and un-pause queues (if the adapter allows that), inspect workers, know which jobs are being run by what worker, checking a specific job or a specific worker...

Default queue tab

In-progress jobs tab

Workers tab

Single job

Single worker

Console helpers, scripting and dealing with big sets of jobs

Besides the UI, Mission Control provides a light console helper to switch between applications and adapters. Some potentially destructive actions aren't exposed via the UI (for example, discarding jobs that aren't failed, although this might change in the future), but you can always perform these from the console if you know very well what you're doing.

It's also possible that you need to deal with very big sets of jobs that are unmanageable via the UI or that you wish to write a script to deal with an incident, some cleanup or some data migration. The console helpers and the querying API with which we've extended Active Job come in handy here.

First, when connecting to the Rails console, you'll see this new message:

 bin/rails c


Type 'jobs_help' to see how to connect to the available job servers to manage jobs

Typing jobs_help, you'll get clear instructions about how to switch between applications and adapters:

>> jobs_help
You can connect to a job server with
  connect_to "<app_id>:<server_id>"

Available job servers:
  * bc4:resque_ashburn
  * bc4:resque_chicago
  * hey:resque
  * hey:solid_queue

And then:

>> connect_to "hey:solid_queue"
Connected to hey:solid_queue

Now you're ready to query and operate over jobs for this adapter via the API. Some examples of queries:


# All jobs
ActiveJob.jobs

# All failed jobs
ActiveJob.jobs.failed

# All pending jobs in some queue
ActiveJob.jobs.pending.where(queue_name: "some_queue")

# All failed jobs of a given class
ActiveJob.jobs.failed.where(job_class_name: "SomeJob")

# All pending jobs of a given class with limit and offset
ActiveJob.jobs.pending.where(job_class_name: "SomeJob").limit(10).offset(5)

# For adapters that support these statuses:
# All scheduled/in-progress/finished jobs of a given class
ActiveJob.jobs.scheduled.where(job_class_name: "SomeJob")
ActiveJob.jobs.in_progress.where(job_class_name: "SomeJob")
ActiveJob.jobs.finished.where(job_class_name: "SomeJob")

# For adapters that support filtering by worker:
# All jobs in progress being run by a given worker
ActiveJob.jobs.in_progress.where(worker_id: 42)

Some examples of bulk operations:

# Retry all the jobs (only possible for failed jobs)
ActiveJob.jobs.failed.retry_all

# Retry all the jobs of a given class (only possible for failed jobs)
ActiveJob.jobs.failed.where(job_class_name: "SomeJob").retry_all

# Discard all failed jobs
ActiveJob.jobs.failed.discard_all

# Discard all pending jobs of a given class
ActiveJob.jobs.pending.where(job_class_name: "SomeJob").discard_all
# Or all pending jobs in a given queue:
ActiveJob.jobs.pending.where(queue_name: "some-queue").discard_all

When performing these bulk operations in the console, a delay of 2 seconds between batches processed will be introduced, set via delay_between_bulk_operation_batches. You can modify it as

MissionControl::Jobs.delay_between_bulk_operation_batches = 5.seconds

Contributing

Thanks for your interest in contributing! To get the app running locally, just run:

bin/setup

This will load a bunch of jobs as seeds.

We have both unit and functional tests and system tests. If you want to run system tests, you'd need to install ChromeDriver. Then, you'll be able to run the tests as:

bin/rails test test/system

License

The gem is available as open source under the terms of the MIT License.