A question about scaling out your backend: In a scenario where you have x pushpin instances and y backend instances. Lets say 4 pushpin instances connected to 2 backend instances. Pushpin1 and pushpin2 are connected to backend1 and pushpin2 og pushpin3 are connected to backend2. Incoming requests are routed to pushpin instances on a round robin basis. If backend1 goes down and is not responding. Is there any way of letting pushpin1 and pushpin2 reconnect with backend2?
hi @larshelg , the best answer is to use a load balancer between the pushpins and backends. so load from pushpin1,2,3,4 would be evenly balanced across backend1,2, and if backend1 goes down then all pushpins send to backend2
pushpin itself does have some fallback capabilities. if a backend hostname has multiple DNS records it will try each until one works. and you can also specify multiple targets for a route (e.g. * backend1:80 backend2:80). but the decisions are made at the time of connect and may involve long timeouts
load balancers typically pre-check each backend in the background, so when a new request arrives it already knows in advance where to forward it so there is no waiting