KIC did not release memory in a timely manner when DP was downsized

[tao12345666333]

Is there an existing issue for this?

• I have searched the existing issues

Current Behavior

I deployed KIC in a new cluster and created an Ingress resource.

When I scaled up the gateway (from 1 to 30) and then scaled it down (from 30 to 1), I found that the memory of KIC was not released promptly.

Expected Behavior

When the number of DP decreases, KIC can correctly release memory.

Steps To Reproduce

1. Install KIC and create an Ingress resource, via https://docs.konghq.com/kubernetes-ingress-controller/latest/get-started/
2. `kubectl top` check the memory of KIC's pod
3. `kubectl scale deployment/kong-gateway -n kong --replicas 30`
4. check the memory again
5. `kubectl scale deployment/kong-gateway -n kong --replicas 1`
6. check the memory again

Kong Ingress Controller version

3.2

Kubernetes version

v1.29

Anything else?

Related to FTI-5975

[czeslavo]

#5068 can be related to this.

[tao12345666333]

Yes, I am fixing #5068, for more details please refer to #5068 (comment)

[tao12345666333]

When I tested the code containing modifications related to #5068, I found that there was not much fluctuation in KIC's memory.

In this test, I ran two replicas of KIC and installed them using Helm. I performed a scale operation on the Gateway, changing the number of replicas from 1 to 20, and then reduced it back to 1 after running for some time.

[czeslavo]

@tao12345666333 Are you still able to reproduce the issue with KIC 3.2 without modifications? I tried reproducing leaks (or #5068 ready list being stuck) but didn't succeed.

[tao12345666333]

Let me try the v3.2 version.

[tao12345666333]

update: The latest code and v3.2 were unable to reproduce the issue.

I will try KIC + Konnect to reproduce this issue.

[randmonkey]

@tao12345666333 Could not reproduce in the latest code, close this.