We have cluster setup without httpcluterservlet and no proxyserver as well.We want to do a setup where if one of the member of the cluster is overloaded it should send out a 503 response and redirect the user to next avaiable server in the cluster. What kind of workmanager we should have and what level to define at. Where to define …at cluster level or server level. What happens to the current sessions in the server if we suspend it. Will the load balancer know that oervloaded server is no longer accepting the sessions and mark it down and kill the sessions in it.
unfortunately there is no way to stop sending request on certain server if it's under overload however there are certain options you can define from console like in case of any server overloaded like immediately restart the server, second you can used different cluster algorithms ( round robin / random / Weight base ) to handle your requests. Work managers you can use to define the priority of your requests and if you want certain threads to be available all the time for some of your specific important critical application.