We have quite a lot of data node here actually but i think something's wrong with our cluster.
of all existing data nodes, only data-1 has the highest CPU usage. it looks like he's working alone on that cluster. you can see too the shards are pretty balance there but when i get into the data-1 i found many primary shards converging on data-1 from many indexes in our cluster
so, is the cpu usage in that node so high because of that? if so, is the use of cpu that big to be reasonable?
How many primary shards for each index? If you have one primary shards per index and ingesting documents intensely to that index whose prinary shard is allocated to Data-1 in chance, the high cpu only in Data-1 is reasonable.
Actually each index has a various count of their primary shard. And about what you said, does that mean the occurrence in data-1 is reasonable because most of primary shards are in data-1? so you mean?