Hello, I have few question about managing rgw multisite deployments. A) zonegroups and zone endpoints can have multiple URLs/addresses. Is it better to pass just one url to load balancer or configure separate address for each RGW daemon in the zone. B) Rook documentation suggest that having more than 1 RGW running replication is bad for performance. I've also seen some IBM blog posts discussing rgw multisite with graphics showing only single RGW running replication traffic. I don't think there are any warnings about this in official documentation. Is this something you consider when planning your multisite deployments? How many RGW dedicated to replication do you usually deploy per zone? https://www.rook.io/docs/rook/v1.17/Storage-Configuration/Object-Storage-RGW... C) Afaik most radosgw-admin command should be executed on ceph cluster with master zonegroup. When I commit changes after these operation period doesn't seem to update on secondary zones. Is this supposed to work like this? D) How are we supposed to keep period on secondary zone in sync? If I need to change secondary zone endpoint I would have always pull the current realm from master zone? 1. pull the latest realm/period from master zone. 2. modifying the secondary zone zone 3. run radosgw-admin period update --commit from secondary zone. E) Do radosgw-admin commands modifying zone on secondary cluster require committing the period afterwards? (for example zone storage class pool change doesn't directly affect zonegroup ) Do they always require rgw restart? Example with enabling compression doesn't commit the period and doesn't mention restarting rgw. https://docs.ceph.com/en/latest/radosgw/compression/ Placement/Storage Classes example mentions that you have to commit period and restart rgw after change. https://docs.ceph.com/en/latest/radosgw/placement/#adding-a-storage-class F) Is it safe to restart RGW daemons during full zone sync? Can it interrupt or negatively affect zone synchronization? Best regards Adam Prycki