Update concept-disaster-recovery.md

Author: vicancy

articles/azure-web-pubsub/concept-disaster-recovery.md

@@ -97,11 +97,14 @@ You'll need to handle such cases at client side to make it transparent to your e
 
 ###  High available architecture for client-client pattern
 
-For client-client pattern, currently it is not yet possible to support a zero-down-time disaster recovery. 
-
-The general principles when handling disaster recovery for client-client pattern are:
-1. Make sure there is only one **primary** Web PubSub service available for all the running app server instances across regions.
-2. Make sure there is no active clients connected to **secondary** Web PubSub services.
+For client-client pattern, currently it is not yet possible to support a zero-down-time disaster recovery. If you have high availability requirements, please consider using client-server pattern, or sending a copy of messages to the server as well.
+
+Clients connected to one Web PubSub service are not yet able to communicate with clients connected to another Web PubSub service using client-client pattern. So when using client-client pattern, the general principles are:
+1. Make sure all the running app server instances across regions are reading to the same configuration.
+2. Make sure the configuration contains only one **primary** Web PubSub service. All the active clients should be connected to this **primary** Web PubSub service.
+3. When the **primary** Web PubSub service is down, the configuration should be updated to promote one and only one **secondary** Web PubSub service endpoint to **primary**.
+4. When the **primary** Web PubSub service is back, the configuration should be updated to change the **primary** endpoint.
+5. Make sure there is no active clients connected to **secondary** Web PubSub services. [Close All Connections](/rest/api/webpubsub/dataplane/web-pub-sub/close-all-connections) could be used to close all the connected clients.
 
 ## How to test a failover