1. start the server, connect to it, set max-connections : /subsystem=undertow/server=default-server/http-listener=default:write-attribute(name=max-connections,value=20) reload 2. deploy attached web archive that enables you to create long-running connection to server: deploy long-running-servlet.war 3. perform request to server like this (duration means how long time request should take): http: //localhost:8080/ long -running-servlet/HeavyProcessing?duration=5000 Ok, so perform requests, eg: for i in {1..300}; do curl http: //localhost:8080/ long -running-servlet/HeavyProcessing?duration=5000 & done and see connections, eg: while true ; do output=$(lsof -i | grep -e "localhost" -e "127.0.0.1" ); echo "${output}" ; echo -n "Connections: " ; echo "${output}" | grep "webcache->.*(ESTABLISHED" | wc -l; echo "---" ; sleep 1; done when using ER2, ER3 and ER4 with max-connections attribute undefined then all connections are processed in parallel - this is just fine. when using ER2 with max-connections attribute set to 20 then processing of created requests is limited apparently - this is just fine. when using ER3 or ER4 with max-connections attribute set to 20 then all connections are processed in parallel again, which is wrong.