Error Handling

Thundering Herd

A problem where a large number of clients simultaneously retry or reconnect after a shared failure, overwhelming the server that has just begun recovering. The name comes from the image of a herd of animals stampeding together. Thundering herd is a natural consequence of synchronized retry schedules — all clients back off for the same duration and then hammer the server at the same moment. Jitter, staggered reconnect windows, and circuit breakers are the primary mitigations.

Protocoles associés

Mentionné dans les guides

Voir aussi