Retries and answering webhooks

How Tamr retries a failed webhook delivery over three days, what turns an endpoint off, and how to de-duplicate and order events.

Answer every delivery with any 2xx status within 15 seconds, and do the work afterwards. Anything else is tried again.

The retry schedule

Anything other than a 2xx (another status, a redirect, a timeout, no connection) is tried again after 5 seconds, 5 minutes, 30 minutes, 2 hours, 5 hours, 10 hours, 14 hours, 20 hours and 24 hours: ten attempts in all, over about three days. A Retry-After header asking for longer is honoured, up to a day.

What turns an endpoint off

  • 410 Gone turns the endpoint off at once.
  • An endpoint that fails every delivery for three days is turned off.

The restaurant's admins are notified. Turning the endpoint back on starts afresh, and the deliveries it missed can be retried from Deliveries in the dashboard.

Duplicates and order

Deliveries can arrive more than once and out of order: a retry can land after a later event.

  • De-duplicate by webhook-id. It is the same on every retry of one event.
  • Order by the payload's own times (endedAt, submittedAt, the envelope's timestamp), never by arrival.
  • Some events are sent again on purpose: a conversation that carries on after it went idle ends again, and a changed rating is a new review.submitted with the same reviewId. Each event's page says which.