Skip to content

Retry storms after an outage

This page covers what “retry storms after an outage” usually means and how to recover.

  • Huge spike in delivery attempts after a customer’s endpoint comes back online.
  • Customer complaining they got the same webhook 12 times in a second.
  • Harbor’s default retry schedule is bursty by design — when a destination comes back, all backlogged retries fire within a short window.
  • No jitter configured on a custom retry policy.
  • Enable per-destination rate limiting (set max_concurrent_deliveries on the destination).
  • For destinations you know are fragile, configure a longer base delay in the retry policy.
  • Coordinate with the customer: ‘expect a burst when you come back online; here’s how to smooth it on your side.’