Suppose I have three nodes A, B, C and three indexes index1, index2, index3, each with one primary shard and one replica shard. If I distribute the shards as follows:
A index1,index2
B index2,index3
C index3,index1
If there is a sudden surge in query requests for index1, causing nodes A and C to reach 100% CPU usage, since shards of index2 and index3 are on node B, this node's CPU load should remain normal. Additionally, with Elasticsearch's Adaptive Replica Selection, when nodes A and C are at full CPU load, query requests for index2 and index3 should be allocated to node B as much as possible. Therefore, index2 and index3 should be able to function close to normal under these circumstances, right?