Sume webhook retries: 10 attempts 30 seconds apart, size your downtime

Sume retries a failed job webhook up to 10 attempts in total, with a fixed delay between attempts that is 30 seconds by default and a 10-second timeout for each attempt. At default spacing that is nine gaps of 30 seconds, so a publisher outage longer than about 4.5 minutes can lose the automatic deliveries; the jobs still finish.

After the attempts are used, recover with POST /v1/jobs/{job_id}/webhook/redeliver (jobs:write) or by polling status_url. Redeliver sends a fresh timestamp and signature and does not use one of the automatic 10.

What are the delivery rules?

From Sume's webhook docs.

How do I find the jobs I missed?

After an outage, list the job ids you submitted that have no stored terminal event, read each status, and redeliver or fetch the result. The sample finds the gaps.

def missing(submitted: list, received: set) -> list:
    return [job_id for job_id in submitted if job_id not in received]


if __name__ == "__main__":
    print(missing(["job_1", "job_2", "job_3"], {"job_1", "job_3"}))

How long can I be down?

Plan for the arithmetic: the 4.5 minutes is a calculation from the documented defaults, not a promise, so treat any deploy that takes the receiver down longer as needing a sweep. The post on a reconcile sweeper shows the loop.