Query Time Performance Degraded drastically after migration

Hi, I recently migrated from my self hosted cluster to Elastic Cloud. The migration came along with an upgrade from ElasticSearch 2.4 to 5.6. Everything seems to work fine despite the fact that the queries on new cluster are much more slower on Elastic Cloud (I'd say 10x slower).

The _reindex Api was used to perform the remote migration. New indices were created on ElasticCloud following this mapping:

Properties of type string and Not analyzed were mapped to keyword.
Other properties of type string were mapped to text.

There are approximately 100M documents over 4 indices. Each index has 5 shards and 1 replica. Same data and settings of previous cluster.

The previous cluster was running on top of 3G of HEAP SIZE. Current ElasticCloud cluster has 8G.

For instance, running this query:

GET my_index/_search
"query": {
"query_string" : {
"query" : "Good OR Awesome OR Great"

I get this result on my self hosted cluster:

"took" : 1389,
"timed_out" : false,
"_shards" : {
"total" : 5,
"successful" : 5,
"failed" : 0
"hits" : {
"total" : 39666020,
"max_score" : 2.1549606,
"hits" : {CENSORED}

And this from Elastic Cloud:

"took": 8838,
"timed_out": false,
"_shards": {
"total": 5,
"successful": 5,
"skipped": 0,
"failed": 0
"hits": {CENSORED}

The hits seems be different, but that I was expecting due change of major versions. But the difference of time is quite painful - and unusable. Elastic Cloud uses SSD.

Any clue or suggestion on how to address this issue?


This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.