Ceph » rgw

As part of tracker 61532 following params were added that would decide how long will the persistent notifications be retried to deliver it to broker in event of failure. If value set is non-zero, after exhaustion the event will be evicted from storage.

rgw_topic_persistency_time_to_live
rgw_topic_persistency_max_retries

As part of desing it was decide these params can be set either by rgw admin (ceph.conf/global value) or by end user when doing Sns::CreateTopic/SetAttribute() (per topic value). And per topic value OVERRIDES the global configuration.

So as part of this tracker, i am trying to propose let the rgw admin decide on OVERRIDES instead of by default user getting the preference to override these values.

• The whole reason for the adding the expiry mechanism was to give the storage admin the ability to expire the messages that would otherwise stay & retried forever if the broker is set in-correctly. And by giving the user value preference we defeat this purpose and user can still set the expiry forever
• Second, these attributes are ceph specific attributes and not AWS standard/compatible, so it makes sense to hide these attributes from user perspective.

To come to middle ground, we could provide a flexibility on the override preference and have Global flag controlled by rgw-admin that decides the override preference. Depending on how we wanna name the flag, value of that flag will decide whether the user value (per topic value) or the global value (ceph.conf value) takes preference.