How do you retry a webhook?
Retrying a webhook means automatically re-attempting a delivery when the destination rejects it or fails to respond — with backoff, per-destination isolation, and a terminal failure surface so nothing dies silently. Slashbin runs retries for every destination automatically, so your API handler never has to.
Slashbin is a webhook ETL and programmable gateway for developers and the AI agents that work alongside them. You own the Golden Model — one canonical schema per topic on each data stream — and every webhook is validated against it, then mapped into the shape each destination needs. Retry, replay, fan-out and the dead-letter queue keep that contract reliable, and the console, the CLI and an agent are three ways to operate the same system.
Retry protects the contract at the delivery step. The Per-Destination Backoff detail below shows a failing destination re-attempted alone, and the Golden Model page shows how the contract those deliveries keep is defined.
Retry vs replay — the distinction
Retry is automatic redelivery after a failure. Slashbin owns the schedule: when a destination returns a retryable error or times out, the gateway backs off and tries again on its own, per destination, until the delivery succeeds or the attempt budget is exhausted.
Replay is manual re-processing of history. You pick a window — an event, a time range, an entire source — and Slashbin re-delivers those stored payloads through your latest transform. See Replay & Retry for the full replay surface.
Webhook retry logic — what good retries require
- Backoff, not tight loops. Immediate retries pile onto a struggling destination and turn a blip into an outage. The retry schedule spaces attempts out so a slow endpoint gets room to recover.
- Per-destination circuit breaking. When one destination starts failing consistently, retries to that destination pause without touching the others. A downed CRM never blocks deliveries to your warehouse.
- Idempotent, no duplicate sends. The gateway acks with the destination before marking a delivery successful. If a retry races an in-flight ack, the state model makes duplicate delivery impossible for the same delivery attempt.
- Terminal failures are visible. When retries exhaust, the delivery is permanently failed for that destination and its delivery status shows it — never a swallowed error in a log line no one reads.
Webhook retry after failure — what happens when the endpoint comes back
When a destination recovers, its backlog drains on its own. Each delivery goes out on its own scheduled retry time, and Slashbin makes no ordering promise between deliveries — so if two versions of the same record are waiting, guard your writes with a timestamp check. One protection you do get on this path: a retry is skipped when a newer delivery for that same record has already succeeded to that same destination. A delivery that exhausted its retries during the outage is permanently failed for that destination and shows that in its delivery status; to send it again, replay the event — it re-runs through the current transform, exactly as the vendor originally sent it.
Because retry is per destination, one destination's recovery doesn't wait for another. Analytics can catch up while your fulfillment failures are still being triaged.
Why hand-rolling retries in your API is a trap
| Approach | What breaks |
|---|---|
| Retry in the webhook handler | The vendor's own timeout fires first. The sender gives up, you retry, and the vendor sends the same webhook again — now you have two in-flight attempts and no way to reconcile them. |
| A cron job that re-sends from a table | You now own a delivery table, a scheduler, a backoff policy, a DLQ surface, and a runbook — plus every destination has to be added to it by hand. |
| Ask the vendor to resend | Not every vendor supports it; those that do rate-limit, batch, or omit events you already acknowledged. |
| Retry in Slashbin | The stored payload, the destination state, and the backoff schedule are already there. Failures retry themselves; a delivery that exhausts its retries is marked permanently failed and stays visible in its delivery status. |
Feature detail
Per-Destination Backoff
Each destination retries on its own schedule — one slow endpoint never delays another.
Circuit Breaking
A destination that fails consistently pauses without affecting deliveries to any other destination.
Visible Terminal Failures
A delivery that exhausts its retries is marked permanently failed for that destination and shows in its delivery status — replay the event to send it again once the endpoint recovers.
No Duplicate Sends
Retries never race in-flight acknowledgements — a successful delivery is not retried, and a retry does not create a second in-flight attempt.
Worked example
Learn how retry and replay work together.
View Setup Guides