The CPU load increases, the CPU utilization decreases, and the Version Map continues to decrease for 30 hours

question1:Why are indicators suddenly like this?
java code ex:Exception message failed to refresh index: user_rank_total

question2:Why is the sharding uneven? node3 is not sharded

It is not clear to me what you are looking for. Please provide more information about the cluster, e.g. version used, configuration, data, indices and shards etc. Please also provide more context around the graphs you have provided and what the perceived issue is.