From 55ebaee4a78edb4524f3324d1cb1a5740c4571f8 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?G=C3=A9rald=20Cro=C3=ABs?= Date: Thu, 13 Mar 2025 09:44:04 +0100 Subject: [PATCH] Clarifies that retry middleware uses TCP, not HTTP status codes --- docs/content/middlewares/http/retry.md | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-) diff --git a/docs/content/middlewares/http/retry.md b/docs/content/middlewares/http/retry.md index 34b4ab695..9ec79c709 100644 --- a/docs/content/middlewares/http/retry.md +++ b/docs/content/middlewares/http/retry.md @@ -12,8 +12,11 @@ Retrying until it Succeeds TODO: add schema --> -The Retry middleware reissues requests a given number of times to a backend server if that server does not reply. -As soon as the server answers, the middleware stops retrying, regardless of the response status. +The Retry middleware reissues requests a given number of times when it cannot contact the backend service. +This applies at the transport level (TCP). +If the service does not respond to the initial connection attempt, the middleware retries. +However, once the service responds, regardless of the HTTP status code, the middleware considers it operational and stops retrying. +This means that the retry mechanism does not handle HTTP errors; it only retries when there is no response at the TCP level. The Retry middleware has an optional configuration to enable an exponential backoff. ## Configuration Examples