When I use FoundationDB, I deploy a three-node service cluster, each node has 9 storage processes, 2 log processes, and 3 stateless processes, the redundancy mode is double, and the proxy is set to 4 multiversion database", and the cluster will not recover after restarting the memory excess process, which did not appear in my previous use, is there a problem with this deployment method, whether it is related to the number of stateless processes, or related to the number of logs configurations, can you give me a deployment suggestion, or a way to avoid this problem.