Doppler

Intro

Doppler helps you schedule HTTP callbacks with retries. Callbacks are HTTP POST and a Flask helper takes care of signing and unsigning payloads.

There are three components to Doppler

1. Flask helper to define webhook endpoitns and and schedule callbacks
2. Queue worker to trigger callback requests at the right time
3. REST service to accept delay requests, cancellations and job introspection

One Liner

Let’s say we have an internal endpoint in Flask that sends users reminders. This is how you would trigger it in 3 hours:

remind_user.delay(60 * 60 * 3, user_id='foo', message='Don’t forget to buy milk')

Accuracy

The Doppler server uses rpqueue and redis to trigger callbacks. I’m not sure what the resolution is, but I wrote this with "minute resolution" as a goal. If you need second or microsecond accuracy look elsewhere.

Use Cases

• Send users notification reminders at a predictable moment in time
• Trigger charges only an hour after a purchase to avoid payback on cancellations
• You get the point...

Example

from flask import Flask, abort
from doppler.ext import Doppler

app = Flask(__name__)
app.config.update(
    SERVER_NAME='localhost:5001',
    SECRET_KEY='foo',
)
doppler = Doppler('http://localhost:5000/')


@doppler.listen('/expire', max_retries=5, retry_delay=10)
def expire(id):
    try:
        Contract.query.get(id).expire()
    except NotSignedError:
        abort(400)  # Will retry 5x with 10 second delays

doppler.register(app)

if __name__ == "__main__":
    app.run(port=5001)

The above Flask app now has an endpoint on /_callbacks/expire.

To schedule a delayed and secure request to this endpoint:

job = expire.delay(10, id='foo')

Canceling Jobs

job = expire.delay(10, id='foo')
job.cancel()

... and querying the state

job = expire.delay(10, id='foo')
job.state == 'early'
time.sleep(10)
job.refresh()  # or doppler.get_job(job.request_id)
job.state == 'done'

Here are the attributes available on jobs

{
  "request_id": "9cbb9962-585a-4240-82a5-ab2537a0fb1b",
  "status": "early",
  "run_at": 1486412420,
  "scheduled_at": 1486412420,
  "last_retry": null,
  "retries_left": 3
}

Security

Note that the endpoint kwargs are not based on URL matching. The doppler.listen decorator takes care of sending and storing callback arguments (using the Werkzeug session cookie implementation). This means the doppler service is treated like a browser and uses your Flask app secret to make sure information originating from your internal server is not shared unencrypted with the doppler server. A public Doppler instance will of course allow anyone to cancel your jobs if they own the job request id’s (does this mean it’s secure if you use HTTPS? I don’t know, I’m not a security expert).