Honeydew ("Honey, do!") is a pluggable job queue and worker pool for Elixir, focused on at-least-once execution.
defmodule MyWorker do
def do_a_thing do
IO.puts "doing a thing!"
end
end
:ok = Honeydew.start_queue(:my_queue)
:ok = Honeydew.start_workers(:my_queue, MyWorker)
:do_a_thing |> Honeydew.async(:my_queue)
# => "doing a thing!"
Isolation
- Jobs are run in isolated one-time-use processes.
- Optionally stores immutable state loaned to each worker (a database connection, for example).
- Initialized Worker
Strong Job Custody
- Jobs don't leave the queue until either they succeed, are explicitly abandoned or are moved to another queue.
- Workers are issued only one job at a time, no batching.
- If a worker crashes while processing a job, the job is reset and a "failure mode" (e.g. abandon, move, retry) is executed. (The default failure mode is to abandon the job.)
- Job Lifecycle
Clusterable Components
- Queues, workers and your enqueuing processes can exist anywhere in the BEAM cluster.
- Global Queues
Plugability
- Queues, workers, dispatch strategies, failure modes and success modes are all plugable with user modules.
- No forced dependency on external queue services.
Batteries Included
- Mnesia Queue, for in-memory/persistence and simple distribution scenarios. (default)
- Ecto Queue, to turn an Ecto schema into its own work queue, using your database.
- Fast In-Memory Queue, for fast processing of recreatable jobs without delay requirements.
- Can optionally heal the cluster after a disconnect or downed node when using a Global Queue.
- Delayed Jobs
- Exponential Retry, even works with Ecto queues!
Easy API
- Jobs are enqueued using
async/3
and you can receive replies withyield/2
, somewhat like Task. - API Overview
- Hex Docs
The Ecto Queue is designed to painlessly turn your Ecto schema into a queue, using your repo as the backing store.
- You don't need to explicitly enqueue jobs, that's handled for you (for example, sending a welcome email when a new User is inserted).
- Eliminates the possibility of your database and work queue becoming out of sync
- As the database is the queue, you don't need to run a separate queue node.
- You get all of the high-availability, consistency and distribution semantics of your chosen database.
Check out the included example project, and its README.
In your mix.exs file:
defp deps do
[{:honeydew, "~> 1.5.0"}]
end
If you're using the Mnesia queue (the default), you'll need tell your release system to include the :mnesia
application, and you'll have to decide how you're going to create your on-disk schema files, which needs to be done while mnesia is not running.
If you use mnesia outside of Honeydew, you'll want to use the :extra_applications
configuration key in your mix.exs file, as well as manually creating your mnesia schema with :mnesia.create_schema(nodes)
in an iex session in production:
def application do
[
extra_applications: [:mnesia]
]
end
Otherwise, if Honeydew is the only user of mnesia, you can let Honeydew manage it by simply using the :included_applications
key instead.
def application do
[
included_applications: [:mnesia]
]
end
- Check out the examples.
- Enqueue jobs with
Honeydew.async/3
, delay jobs by passingdelay_secs: <integer>
. - Receive responses with
Honeydew.yield/2
. - Emit job progress with
progress/1
- Queue/Worker status with
Honeydew.status/1
- Suspend and resume with
Honeydew.suspend/1
andHoneydew.resume/1
- List jobs with
Honeydew.filter/2
- Move jobs with
Honeydew.move/2
- Cancel jobs with
Honeydew.cancel/2
The rest of the README is broken out into slightly more digestible sections.
Also, check out the README files included with each of the examples.
It's worth keeping abreast with the CHANGELOG