proudly made by Chaps
Gush is a parallel workflow runner using only Redis as storage and ActiveJob for scheduling and executing jobs.
Gush relies on directed acyclic graphs to store dependencies, see Parallelizing Operations With Dependencies by Stephen Toub to learn more about this method.
This README is about the 1.0.0
version, which has breaking changes compared to < 1.0.0 versions. See here for 0.4.1 documentation.
gem 'gush', '~> 1.0.0'
When using Gush and its CLI commands you need a Gushfile
in the root directory.
Gushfile
should require all your workflows and jobs.
For RoR it is enough to require the full environment:
require_relative './config/environment.rb'
and make sure your jobs and workflows are correctly loaded by adding their directories to autoload_paths, inside config/application.rb
:
config.autoload_paths += ["#{Rails.root}/app/jobs", "#{Rails.root}/app/workflows"]
Simply require any jobs and workflows manually in Gushfile
:
require_relative 'lib/workflows/example_workflow.rb'
require_relative 'lib/jobs/some_job.rb'
require_relative 'lib/jobs/some_other_job.rb'
The DSL for defining jobs consists of a single run
method.
Here is a complete example of a workflow you can create:
# app/workflows/sample_workflow.rb
class SampleWorkflow < Gush::Workflow
def configure(url_to_fetch_from)
run FetchJob1, params: { url: url_to_fetch_from }
run FetchJob2, params: { some_flag: true, url: 'http://url.com' }
run PersistJob1, after: FetchJob1
run PersistJob2, after: FetchJob2
run Normalize,
after: [PersistJob1, PersistJob2],
before: Index
run Index
end
end
and this is how the graph will look like:
Let's start with the simplest workflow possible, consisting of a single job:
class SimpleWorkflow < Gush::Workflow
def configure
run DownloadJob
end
end
Of course having a workflow with only a single job does not make sense, so it's time to define dependencies:
class SimpleWorkflow < Gush::Workflow
def configure
run DownloadJob
run SaveJob, after: DownloadJob
end
end
We just told Gush to execute SaveJob
right after DownloadJob
finishes successfully.
But what if your job must have multiple dependencies? That's easy, just provide an array to the after
attribute:
class SimpleWorkflow < Gush::Workflow
def configure
run FirstDownloadJob
run SecondDownloadJob
run SaveJob, after: [FirstDownloadJob, SecondDownloadJob]
end
end
Now SaveJob
will only execute after both its parents finish without errors.
With this simple syntax you can build any complex workflows you can imagine!
run
method also accepts before:
attribute to define the opposite association. So we can write the same workflow as above, but like this:
class SimpleWorkflow < Gush::Workflow
def configure
run FirstDownloadJob, before: SaveJob
run SecondDownloadJob, before: SaveJob
run SaveJob
end
end
You can use whatever way you find more readable or even both at once :)
Workflows can accept any primitive arguments in their constructor, which then will be available in your
configure
method.
Let's assume we are writing a book publishing workflow which needs to know where the PDF of the book is and under what ISBN it will be released:
class PublishBookWorkflow < Gush::Workflow
def configure(url, isbn)
run FetchBook, params: { url: url }
run PublishBook, params: { book_isbn: isbn }, after: FetchBook
end
end
and then create your workflow with those arguments:
PublishBookWorkflow.create("http://url.com/book.pdf", "978-0470081204")
and that's basically it for defining workflows, see below on how to define jobs:
The simplest job is a class inheriting from Gush::Job
and responding to perform
method. Much like any other ActiveJob class.
class FetchBook < Gush::Job
def perform
# do some fetching from remote APIs
end
end
But what about those params we passed in the previous step?
To do that, simply provide a params:
attribute with a hash of parameters you'd like to have available inside the perform
method of the job.
So, inside workflow:
(...)
run FetchBook, params: {url: "http://url.com/book.pdf"}
(...)
and within the job we can access them like this:
class FetchBook < Gush::Job
def perform
# you can access `params` method here, for example:
params #=> {url: "http://url.com/book.pdf"}
end
end
Now that we have defined our workflow and its jobs, we can use it:
Important: The command to start background workers depends on the backend you chose for ActiveJob. For example, in case of Sidekiq this would be:
bundle exec sidekiq -q gush
Click here to see backends section in official ActiveJob documentation about configuring backends
Hint: gush uses gush
queue name by default. Keep that in mind, because some backends (like Sidekiq) will only run jobs from explicitly stated queues.
flow = PublishBookWorkflow.create("http://url.com/book.pdf", "978-0470081204")
flow.start!
Now Gush will start processing jobs in the background using ActiveJob and your chosen backend.
flow.reload
flow.status
#=> :running|:finished|:failed
reload
is needed to see the latest status, since workflows are updated asynchronously.
Gush offers a useful tool to pass results of a job to its dependencies, so they can act differently.
Example:
Let's assume you have two jobs, DownloadVideo
, EncodeVideo
.
The latter needs to know where the first one saved the file to be able to open it.
class DownloadVideo < Gush::Job
def perform
downloader = VideoDownloader.fetch("http://youtube.com/?v=someytvideo")
output(downloader.file_path)
end
end
output
method is used to ouput data from the job to all dependant jobs.
Now, since DownloadVideo
finished and its dependant job EncodeVideo
started, we can access that payload inside it:
class EncodeVideo < Gush::Job
def perform
video_path = payloads.first[:output]
end
end
payloads
is an array containing outputs from all ancestor jobs. So for our EncodeVide
job from above, the array will look like:
[
{
id: "DownloadVideo-41bfb730-b49f-42ac-a808-156327989294" # unique id of the ancestor job
class: "DownloadVideo",
output: "https://s3.amazonaws.com/somebucket/downloaded-file.mp4" #the payload returned by DownloadVideo job using `output()` method
}
]
Note: Keep in mind that payloads can only contain data which can be serialized as JSON, because that's how Gush stores them internally.
There might be a case when you have to construct the workflow dynamically depending on the input.
As an example, let's write a workflow which accepts an array of users and has to send an email to each one. Additionally after it sends the e-mail to every user, it also has to notify the admin about finishing.
class NotifyWorkflow < Gush::Workflow
def configure(user_ids)
notification_jobs = user_ids.map do |user_id|
run NotificationJob, params: {user_id: user_id}
end
run AdminNotificationJob, after: notification_jobs
end
end
We can achieve that because run
method returns the id of the created job, which we can use for chaining dependencies.
Now, when we create the workflow like this:
flow = NotifyWorkflow.create([54, 21, 24, 154, 65]) # 5 user ids as an argument
it will generate a workflow with 5 NotificationJob
s and one AdminNotificationJob
which will depend on all of them:
-
of a specific workflow:
bundle exec gush show <workflow_id>
-
of all created workflows:
bundle exec gush list
This requires that you have imagemagick installed on your computer:
bundle exec gush viz <NameOfTheWorkflow>
Running NotifyWorkflow.create
inserts multiple keys into Redis every time it is ran. This data might be useful for analysis but at a certain point it can be purged via Redis TTL. By default gush and Redis will keep keys forever. To configure expiration you need to 2 things. Create initializer (specify config.ttl in seconds, be different per environment).
# config/initializers/gush.rb
Gush.configure do |config|
config.redis_url = "redis://localhost:6379"
config.concurrency = 5
config.ttl = 3600*24*7
end
And you need to call flow.expire!
(optionally passing custom TTL value overriding config.ttl
). This gives you control whether to expire data for specific workflow. Best NOT to set TTL to be too short (like minutes) but about a week in length. And you can run Client.expire_workflow
and Client.expire_job
passing appropriate IDs and TTL (pass -1 to NOT expire) values.
Since we do not know how long our workflow execution will take we might want to avoid starting the next scheduled workflow iteration while the current one with same class is still running. Long term this could be moved into core library, perhaps Workflow.find_by_class(klass)
# config/initializers/gush.rb
GUSH_CLIENT = Gush::Client.new
# call this method before NotifyWorkflow.create
def find_by_class klass
GUSH_CLIENT.all_workflows.each do |flow|
return true if flow.to_hash[:name] == klass && flow.running?
end
return false
end
- Fork it ( http://github.com/chaps-io/gush/fork )
- Create your feature branch (
git checkout -b my-new-feature
) - Commit your changes (
git commit -am 'Add some feature'
) - Push to the branch (
git push origin my-new-feature
) - Create new Pull Request