A load balancer helps to monitor the health of the backend instances. If one of the instances is unavailable, it would be marked as unavailable. Similarly, if the instances are healthy, it would be available to serve the requests. Traffic would be routed only to available healthy instances. This provides a fault tolerant application so the traffic reaching the application is not affected when we have unavailable instances in the backend. However, if none of your systems are unavailable in the backend to serve requests, the load balancer marks all the instances as unhealthy and users will be affected by an unavailable application.