I want to know why the cluster will undergo a brief restart when there is a memory warning for the storage server’s MVCC. Is there any way to avoid this issue?
At the same time, the memory usage of one node is very low about 45%, while the memory usage of other nodes is around 80%.We use 9 storage processes, 1 log process, and 3 stateless processes on a single node, in this case, a little more pressure will cause the cluster to enter the “storage server performance” and become storage server mvcc memory after a long duration; But I personally feel that the storage drop rate should be faster than the log write rate, so what is the reason for this phenomenon?