I’m trying to install a new cluster on ECK (version 8.8.1) and I’m having latency issues (~300ms).
We already have a similar cluster working in version 6.5.3 (EC2) and for the same query timing is good (~20ms)
The query structure has a nested query and inner_hits part:
"inner_hits": {
"size": 100,
"_source": false,
"docvalue_fields": [
"products.code"
]
}
When running the profiling API i’m getting the bottleneck is in the InnerHitsPhase :
{
"type": "InnerHitsPhase",
"description": "",
"time_in_nanos": 58284286,
"breakdown": {
"process_count": 73,
"process": 58283846,
"next_reader": 440,
"next_reader_count": 1
}
In terms of resources - There is no stat that indicates a limitation
Is there a configuration that can influence the inner_hits phase timing? (edited)