Overview

The purpose of this test set is to test browse/subject/instances API. Measure performance, find possible issues, bottlenecks.

MSEARCH-293. Test performance of subject browsing.

Summary

1 user test ended without errors. Average response time 5.509 s. Resources usage is hardly visible on mod-search during test (most of work were done on OpenSearch end). OpenSearch CPU usage reached 70%.
5 users test ended without errors. Average response time 7.918s. Resources usage is hardly visible on mod-search during test (most of work were done on OpenSearch end). OpenSearch CPU usage reached 100%.
10 users ended without errors as well. Average response time is 13.357 s. OpenSearch CPU usage reached 100% immediately after test start.
1-100 users Ramp-up.
- first error appeared on 16 users;
- on 35 users mod-search becomes unstable. * see more details below.
OpenSearch v 1.2 working much slower than Elasticsearch 7.10. See details in last tests.

mod-search 1.6.4
OpenSearch service
- nodes type r6g.large.search (X4) * (4 data nodes 3 dedicated master nodes);
- version OpenSearch 1.2 (latest);
kafka 2.8.0

users	Average	75%	Errors
1	5.509	6.164	None

Note: This test is not visible on most charts, except CPU usage from OpenSearch service.

Note: OpenSearch CPU usage reached ±70% during test.

*Rest of a charts does not contain any valuable information.

users	Average	75%	Errors
5	7.918	9.085	None

users	Average	75%	Errors
10	13.357	15.568	none

Notable observations:

First errors starts at 16 users;
On 35 users system becomes unstable and containers often failed (See CPU chart);
After 35 users - error rate reached 91%.

ERRORS:

On client side - HTTP 500 Internal Server Error. If the issue persists, please report it to EBSCO Connect.
On server side: ERROR andlerExecutionChain HandlerInterceptor.afterCompletion threw exception
java.lang.NullPointerException: null.