Skip to content

NodeNeighborLinkUpdater memory usage regression #39499

@marseel

Description

@marseel

Is there an existing issue for this?

  • I have searched the existing issues

Version

equal or higher than v1.17.3 and lower than v1.18.0

What happened?

Memory regression is happening on main.
Since #39001 was merged, memory usage of cilium-agent increased by 3-4% in scale-tests. Most of the memory increase is coming from health-reporting:

Image

It seems fairly excessive to report health per each node and each service backed:

hr := sc.NewScope(strings.ReplaceAll(e.node.Name, ".", "-"))

Proposal:
Instead of reporting health per each backend/node, let's report health one time with number of refreshed and failed entries.

How can we reproduce the issue?

CI/scale-test

Cilium Version

main

Kernel Version

N/A

Kubernetes Version

N/A

Regression

yes

Sysdump

Pprofs in zip file

files.zip

Relevant log output

Anything else?

No response

Cilium Users Document

  • Are you a user of Cilium? Please add yourself to the Users doc

Code of Conduct

  • I agree to follow this project's Code of Conduct

Metadata

Metadata

Assignees

No one assigned

    Labels

    area/agentCilium agent related.kind/bugThis is a bug in the Cilium logic.kind/regressionThis functionality worked fine before, but was broken in a newer release of Cilium.release-blocker/1.18This issue will prevent the release of the next version of Cilium.sig/scalabilityImpacts how well Cilium handles a high rate of events or churn.

    Type

    No type

    Projects

    Status

    Done

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions