Memory usage doubled after upgrading from v0.9.8 to v0.9.9

[larsgerber]

Bug description

After upgrading the Capsule Proxy from v0.9.8 to v0.9.9 the memory usage doubled. We updated to latest first but then downgraded until we found the version.

The issue only happens on clusters with +150 CRD's and both proxy feature gates enabled.

How to reproduce

I was able to reproduce the issue in a local Kind cluster:
https://github.com/larsgerber/kind/tree/main/capsule/memory

Expected behavior

Memory usage should not increase dramatically without further notice.

Additional context

• Capsule-Proxy version: capsule-proxy:v0.9.9
• Helm Chart version: capsule-proxy-0.9.9
• Kubernetes version: v1.34.0

[oliverbaehler]

Did you enable the webhooks?

[larsgerber]

I'm not sure which option you mean. It's the default installation with the values from https://github.com/larsgerber/kind/blob/main/capsule/memory/values.yaml

options:
  logLevel: 7
  extraArgs:
    - --feature-gates=ProxyClusterScoped=true
    - --feature-gates=ProxyAllNamespaced=true
resources:
  requests:
    cpu: 200m
    memory: 512Mi
  limits:
    cpu: 1000m
    memory: 1Gi

[oliverbaehler]

capsule-proxy/charts/capsule-proxy/values.yaml

Line 334 in a13d7d9

enabled: false

[larsgerber]

No, we don't use the webhooks at the moment.

[oliverbaehler]

Could you try enabling them in your memory test and verify if usage decreases or not? It could help

[larsgerber]

I tested this in my lab.
It made no difference. Since I had to install cert-manager for the MutatingWebhookConfiguration, I switched to using the ArgoCD Application resource as a dummy CRD. (Certificate had ~800 lines, while Application has ~6700.) The memory usage increased from 150 Mi to 600 Mi.

I strongly suspect this is related to the CRD watcher changes.

[github-actions]

This pull request has been automatically closed because it has been inactive for more than 60 days. Please reopen if you still intend to submit this pull request.