From dbed32a7b395a2f56d67b894e8560279a5ee7ab5 Mon Sep 17 00:00:00 2001 From: Piotr Parczewski Date: Mon, 31 Jul 2023 16:42:26 +0200 Subject: [PATCH] doc/radosgw: correct emphasis in rate limit section Signed-off-by: Piotr Parczewski (cherry picked from commit 50bfbab6d9a30b3ba6ec4f4c8df9059298c28a7f) --- doc/radosgw/admin.rst | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/doc/radosgw/admin.rst b/doc/radosgw/admin.rst index 604b5c156103e..9aaa05671d2e7 100644 --- a/doc/radosgw/admin.rst +++ b/doc/radosgw/admin.rst @@ -484,8 +484,8 @@ Every Object Gateway tracks per user and bucket metrics separately, these metric That means that the desired limits configured should be divide by the number of active Object Gateways. For example, if userA should be limited by 10 ops per minute and there are 2 Object Gateways in the cluster, the limit over userA should be 5 (10 ops per minute / 2 RGWs). -if the requests are ``not`` balanced between RGWs, the rate limit may be underutilized. -For example, if the ops limit is 5 and there are 2 RGWs, ``but`` the Load Balancer send load only to one of those RGWs, +If the requests are **not** balanced between RGWs, the rate limit may be underutilized. +For example, if the ops limit is 5 and there are 2 RGWs, **but** the Load Balancer send load only to one of those RGWs, The effective limit would be 5 ops, because this limit is enforced per RGW. If there is a limit reached for bucket not for user or vice versa the request would be cancelled as well. The bandwidth counting happens after the request is being accepted, as a result, even if in the middle of the request the bucket/user has reached its bandwidth limit this request will proceed. -- 2.39.5