Webhook alert delivery: what happens when your endpoint is slow or down
Short answer: point your platform's webhook at your HoldAlarms URL. HoldAlarms answers the sender, holds the alert and delivers it to your endpoint. HoldAlarms retries up to 5 times. Retries finish within about 3 minutes, or up to about 6 with Retry-After. Permanent errors (400, 401, 403, 410, 413, 414, 415) aren't retried.
Who this is for
Anyone whose alerts come from a platform that can send an HTTP webhook (monitoring tools, form and payment tools, charting platforms such as TradingView, or your own scripts) and who needs those alerts to reach a server that is sometimes slow, restarting or briefly offline.
How a delivery works
- The sender posts the alert to your HoldAlarms URL.
- HoldAlarms replies to the sender and queues the alert.
- HoldAlarms forwards the alert to your endpoint. Your endpoint gets 10 seconds per attempt.
- On a timeout, network error or 5xx reply, HoldAlarms tries again, 2 to 60 seconds apart.
- If a delivery fails for good, HoldAlarms parks it in the dead-letter queue and emails you; failure emails are grouped and capped at up to 10 a day.
Make your endpoint safe to call twice
A retry can repeat an alert your endpoint already accepted, for example when its reply is lost. Give each alert an id, store the ids you have handled and ignore repeats. Identical alerts from the same account within 20 seconds are forwarded once.
What HoldAlarms doesn't do
It adds a hop, so it doesn't cut latency. A delivered alert means your endpoint accepted it, nothing more.