I upgraded our cluster from 5.6 to 6.2.4 and search performance went down, especially in aggregations, which we use quite heavily. This is how the cluster monitoring data looks like:
As you can see, Heap and GC raised up, I can also see messages about long GC in cluster logs:
[gc] overhead, spent [4.1s] collecting in the last [4.2s]
We have cluster of 8 nodes, each with 16GB RAM with 8GB heap.
Any ideas what to look for? thank you