Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Istio HPA tuning #810

Closed
4 tasks done
strekm opened this issue May 9, 2024 · 1 comment
Closed
4 tasks done

Istio HPA tuning #810

strekm opened this issue May 9, 2024 · 1 comment
Assignees
Labels
area/service-mesh Issues or PRs related to service-mesh kind/feature Categorizes issue or PR as related to a new feature.

Comments

@strekm
Copy link
Contributor

strekm commented May 9, 2024

Description

Remove memory metric from Istio HPA for ingress gateway for regular profile. For light profile set max replicas for ingress gateway to 1 and disable autoscale. Execute performance tests to rule out OOM.

Reasons

DoD:

  • [ ] Provide unit and integration tests.
  • Provide documentation.
  • Verify if the solution works for both open-source Kyma and SAP BTP, Kyma runtime.
  • [ ] If you changed the resource limits, explain why it was needed.
  • If the default configuration of Istio Operator has been changed, you performed a manual upgrade test to verify that the change can be rolled out correctly.
  • [ ] Verify that your contributions don't decrease code coverage. If they do, explain why this is the case.
  • Add release notes.~~

Attachments



PR

@strekm strekm added kind/feature Categorizes issue or PR as related to a new feature. area/service-mesh Issues or PRs related to service-mesh labels May 9, 2024
@barchw barchw self-assigned this May 13, 2024
@barchw
Copy link
Contributor

barchw commented May 13, 2024

Executed performance tests for scaling based on CPU + Memory vs CPU only. A moderate increase in RTT is seen with P(95) time getting 40-80 ms slower.

No significant raise in memory usage was seen, and the memory level stayed very far from the set limit, meaning that there should be no risk of OOMKills happening.

On light installation, when the replicaNumber is limited to 1, the pod get's CPU bound before it could get up to the memory limit, as CPU rises significantly faster.

I limited maxreplicas from 5 to 1 instead of disabling autoscaling to make sure that configuration from Istio CR is still possible

@barchw barchw removed their assignment May 14, 2024
@triffer triffer self-assigned this May 15, 2024
@barchw barchw assigned barchw and unassigned triffer May 21, 2024
@strekm strekm closed this as completed Jun 7, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
area/service-mesh Issues or PRs related to service-mesh kind/feature Categorizes issue or PR as related to a new feature.
Projects
None yet
Development

No branches or pull requests

3 participants