Elastic search spring JPA saveAll() + performance issue - elasticsearch

Question

I am using spring boot 2.2 version and elastic search 6.4 version
And using spring data elastic search which extends CurdRepository
Problem is saveAll() on elastic search spring data jpa is taking 30 sec on the first iteration to save 500 records and after some iterations it is 14 or 15 sec to save 500 records which is more time and this is happening when the object is processed from cache but the same 500 records took 1 sec when it is fetched from elastic search and saved.
Elastic search document has more columns with nested objects
Elastic search document
#Document(indexname = "example", type = "example1", shards = 1, replicas =0, refreshInterval = -1)
public class Example1 {
...
private Example2 example2;
private Example3 example3;
private List<Example6> example6List;
...
}
#Document(indexname = "example", type = "example2", shards = 1, replicas =0, refreshInterval = -1)
public class Example1 {
...
private Example4 example4;
...
}
#Document(indexname = "example", type = "example3", shards = 1, replicas =0, refreshInterval = -1)
public class Example1 {
...
private Example5 example5;
private List<Example3> example3List;
}
Elastic search repository class
public interface Example1Repository extends ElasticSearchRepository {
}
Service class
#RequiredArgsConstructor
public class ExampleService {
private final Example1Repository example1Respository;
public void exampleMethod1() {
//read data from cache
// process the logic and save the list
// save to elastic search
example1Repository.saveAll(list); <-- this is taking 30 sec for 500 records to save
}
public void exampleMethod2() {
//read data from cache
//process the logic and save the list
//read data from **elastic search** and iterate and set the values from above list
//save to elastic search
example1Repository.saveAll(list); <-- this is taking 1 sec for 500 records to save
}
}
How to improve the performance in exampleMethod1() approach ? or what is causing performance issue ?
Thank you

Answer 1

Spring Boot 2.2 means Spring Data Elasticsearch 3.2, both versions are out of maintenance quite a while.
From a Spring Data Elasticsearch both calls do the same: Create a bulkindex request, send that to Elasticsearch and then do a refresh.
The difference in these requests might be - and I cannot see this from the info you provide - that the first time there is no id set in the objects.
I don't know if Elasticsearch is faster on a bulk index when there already are ids in the document. But you should consider the information about bulk requests from the Elasticsearch documentation for the current version.
Have you checked if there are any relevant messages in the Elasticsearch log?