Executed a query in ES 1.4.4 and 2.4.0 the results shows 3x times performance degrade in 2.4.0 . Both 1.4.4 and 2.4.0 Hardware configurations and Shard configurations are same. Below is the query. Not sure missing any tuning configurations specific to 2.4.0
{
"from" : 0,
"size" : 200,
"query" : {
"bool" : {
"must" : {
"query_string" : {
"query" : "category:vehicles"
}
}
}
},
"_source" : {
"includes" : [ "@type", "name", "model", "year"],
"excludes" : [ ]
},
"aggregations" : {
"results" : {
"terms" : {
"field" : "model",
"size" : 200
},
"aggregations" : {
"hits" : {
"top_hits" : {
"size" : 20,
"_source" : {
"includes" : [ "@type", "name", "model", "year"],
"excludes" : [ ]
},
"highlight" : {
"require_field_match" : false,
"fields" : {
"*" : { }
}
}
}
}
}
}
},
"highlight" : {
"require_field_match" : false,
"fields" : {
"*" : { }
}
}
}
Tried the following optimization techniques no luck.
Tried mapping with doc value true for all fields
Tried mapping doc value false for all fields
"collect_mode" algorithm set as Breadth first
"collect_mode" algorithm set as Depth first
With global_ordinals_low_cardinality enabled
With eager loading of fields enabled & disabled
Profiling the query to debug time taken for each part
Increased query caching of ES server to upto 30%
Scanning ES server for swap memory
Monitoring ES server's CPU, Heap size and RAM
Increased heap size from 50% of RAM upto 75%
Tried using execution hint
Tried executing query directly in ES server (to rule out n/w issues)
Tried executing query in a single ES server instead of a cluster