[BUG] throughput degradation query test

[timofeevmd]

OS and Environment

linux, k8s

GIT commit hash

a62640d8a

Minimum working example / Steps to reproduce

java-sdk version: e23b548 iroha2-perf: 94b7855

test scenario:

queryPostFindAssetById 
queryPostFindAccountById 
queryPostFindAssetQuantityById

The maximum throughput was 1633 RPS without the PARCA plugin and 3041 RPS with the PARCA plugin connected. After reaching peak values, performance decreases.

1. opensearch logs

2. parca profile:

• diff 2_minuts_before_decreases/2_min_after_throughput_peak profile_moreThen_3000_tps.pb.gz
• full profile profile_full_queries_more_then_3000_tps.pb.gz

3. full performance report (test 9, test 10)

config for 9 test: OUTSIDE test with PARCA PROFILER, 5 peer, iroha2_loglevel: DEBUG

config for 10 test: OUTSIDE test, 5 peer, iroha2_loglevel: DEBUG

./mvnw gatling:test -DtargetURL=https://iroha2.test.tachi.soramitsu.co.jp/ -DremoteLogin=i -DremotePassword= -DconcurrentUsers=50 -Dtimes=100 -DstageDuration=18 -DstartingFrom=0 -DrampDuration=18 -DmaxDuration=1800

Actual result

performance decreases

Expected result

he load continues to increase

Logs

Log contents

look at "step for reproduce"

Who can help to reproduce?

@timofeevmd

Notes

No response