Naden,
Atlas HA deployment work on Active-Passive mode i.e. only one instance of Atlas will be active anytime, requests to other instances will be redirected to the active instance. Load-balancer to front Atlas instances may not work, as redirect from a passive instance will directly point to active instance – bye-passing the load-balancer. Atlas server is not stateless. It caches type-definitions in-memory and only one instance is expected to process Kafka notifications (from hooks). When an active instance goes down, one of the passive instances will automatically become active and initialize itself with type-definitions from the store and start processing Kafka notifications. Hope this helps. Madhan From: Naden <[email protected]> Reply-To: "[email protected]" <[email protected]> Date: Wednesday, October 31, 2018 at 12:57 PM To: "[email protected]" <[email protected]> Subject: Horizontal Scalability Hey guys, I am looking to run multiple Atlas instances behind a load balancer all pointing to the same backend. Is Atlas stateless ? And if I disable the DB cache does that change anything ? Cheers, Naden
