> ## Documentation Index
> Fetch the complete documentation index at: https://docs.hooksnode.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Retries

> How hooksnode retries failed deliveries.

A delivery fails when the destination answers with anything other than `2xx`, or does not answer in 15 seconds. hooksnode then tries again.

## The schedule

* Up to **25 attempts** for each destination.
* The wait between attempts grows with each try. It never goes over **1 hour**.
* All 25 attempts take about **17.5 hours**. A destination that is down overnight still gets its events.

Each destination retries on its own. A failing CRM does not slow your ledger.

After the last attempt fails, the delivery is `dead`. hooksnode does not try again by itself. You can [retry it](/delivery/replay).

<Note>A URL that the security guard blocks, such as a private address, is not retried.</Note>

## Failure alerts

When a destination starts to fail, hooksnode emails your team. See [Alerts and summaries](/monitoring/alerts).

## If the server stops

hooksnode saves each event before it queues it. Every 2 minutes, a job finds events that stayed in `received` for more than 90 seconds and queues them again. A crash between the save and the queue does not lose events.

## Make your endpoint retry-safe

Retries mean that a destination can get the same event more than once. For example, your endpoint did the work but timed out before it answered. See [Idempotency](/delivery/idempotency).

Answer fast. Do the slow work after you answer `200`, for example in a job queue.