Runtime: quarkus

Persistance: postgresql

Prop: kogito.persistence.optimistic.lock=true

Process (each sequntial task is executing async by jobs service):

We use a simple load test for Jmeter to send POST requests, the frequency of which gradually increases (over 5 minutes) up to 10 tps.

An example of a load script for Jmeter (for Jmeter to work correctly, you will need to install the plugin Concurrency Thread Group - https://jmeter-plugins.org/wiki/ConcurrencyThreadGroup/):

kogito_service_step_async.jmx

api_http_response_code_total{artifactId="kogito-bpmn-async",endpoint="management/jobs/sequential/instances/d9a2be29-1b2e-49b3-a5db-364bc1314044/timers/ASYNC:00152d97-0edc-4515-b670-eb087cfe6a98",identifier="200",version="1.21.0.Final",} 1.0

api_http_response_code_total{artifactId="kogito-bpmn-async",endpoint="sequential",identifier="201",version="1.21.0.Final",} 20174.0

api_http_response_code_total{artifactId="kogito-bpmn-async",endpoint="NOT FOUND",identifier="404",version="1.21.0.Final",} 2.0

starts with "api_http_response_code_total" count 140521

artifact ids size 140521
artifact ids distinct 1
endpoints size 140521
endpoints distinct 140521
ids 1 size 140519
ids 1 distinct 20147
ids 2 size 140519
ids 2 distinct 140519

Since endpoints are not removed from memory, this problem is one of the causes of memory leaks.