Elasticsearch7.x have serious performance problems

Elasticsearch7.x have serious performance problems.
We have a indexes set which have total 3800M doc, 5000G bytes.
1. Searching performance
We run a lot of queries against Elasticsearch6.2.1 and Elasticsearch7.9.2 cluster.
The average took time on Elasticsearch6.2.1 is _TOOK_AVG62.
The average took time on Elasticsearch7.9.2 is _TOOK_AVG79.
_TOOK_AVG79 / _TOOK_AVG62 = 2.6
Same query takes longer on Elasticsearch7.9.2.

2. Focemerging performance
Our test cluster is 20 data nodes, 190 primary shards, 1 replica.
When we run the focemerge on the ElasticSearch6.2.1 cluster with max_num_segments=1, it will take 5.5 hours.
When we run the focemerge on the ElasticSearch7.9.2 cluster with max_num_segments=1, it will take 17.5 hours.

A 5TB index is pretty large. What sort of data is this?

That's 17 milliseconds difference though. I understand it's larger, but that's still pretty small.

They are user shopping data.

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