keycloak / keycloak-benchmark

Keycloak Benchmark
https://www.keycloak.org/keycloak-benchmark/
Apache License 2.0
125 stars 68 forks source link

Increased GC overhead in the continuous performance tests #827

Closed kami619 closed 1 month ago

kami619 commented 1 month ago

Describe the bug

Increased GC overhead in the continuous performance tests

Version

Keycloak 25

Expected behavior

We want the GC behavior to be consistent.

Actual behavior

In some cases the CPU utilization of one of the pods gets skewed in terms of utilization. While the JVM reports 2.5% overhead, the overall CPU usage of the pod doubles.

How to Reproduce?

TBA

Anything else?

No response

ahus1 commented 1 month ago

With https://github.com/keycloak/keycloak/issues/29939 being resolved, IMHO this issue can be closed.

kami619 commented 1 month ago

Correct @ahus1 I think with https://github.com/keycloak/keycloak/issues/29939 resolved, we could close this one out. I have created a new Epic #839 to look into other possible JVM GC controls for G1GC and group them with https://github.com/keycloak/keycloak-benchmark/issues/781 for the next series of tests and documenting the run-book, metrics of relevance.