From d12ba11741dc749bcce315bf467078595fa95b24 Mon Sep 17 00:00:00 2001 From: Yuval Lifshitz Date: Wed, 14 Aug 2024 11:02:09 +0000 Subject: doc/rgw/notification: persistent notification queue full behavior Fixes: https://tracker.ceph.com/issues/50610 Signed-off-by: Yuval Lifshitz --- doc/radosgw/notifications.rst | 10 ++++++++-- 1 file changed, 8 insertions(+), 2 deletions(-) (limited to 'doc/radosgw/notifications.rst') diff --git a/doc/radosgw/notifications.rst b/doc/radosgw/notifications.rst index 76e8209bd28..2bbd2c52edf 100644 --- a/doc/radosgw/notifications.rst +++ b/doc/radosgw/notifications.rst @@ -61,9 +61,15 @@ Asynchronous Notifications ~~~~~~~~~~~~~~~~~~~~~~~~~~ Notifications can be sent asynchronously. They are committed into persistent -storage and then asynchronously sent to the topic's configured endpoint. In -this case, the only latency added to the original operation is the latency +storage and then asynchronously sent to the topic's configured endpoint. +The notification will be committed to persistent storage only if the triggering +operation was successful. +In this case, the only latency added to the original operation is the latency added when the notification is committed to persistent storage. +If the endpoint of the topic to which the notification is sent is not available for a long +period of time, the persistent storage allocated for this topic will eventually fill up. +When this happens the triggering operations will fail with ``503 Service Unavailable``, +which tells the client that it may retry later. .. note:: If the notification fails with an error, cannot be delivered, or times out, it is retried until it is successfully acknowledged. -- cgit v1.2.3