[proposal] Add E2E tests for koordlet

[saintube]

Background

What is your proposal:

Currently, the end-to-end test for the koordlet is missing. Though the basic metrics reporting and batch overcommitment are tested in the slo-controller's E2E, we expect to check the QoS strategies provided by the koordlet like BECPUSuppress, BEMemoryEvict, BECPUEvict, CPUBurst, CPU QoS/Memory QoS, and so on.

Why is this needed:

Is there a suggested solution, if so, please add it:

Since #1573 and #1688, the E2E workflows are enabled on the CI pipeline of GitHub Actions. We can easily add new E2E tests in test/e2e and .github/workflows/e2e-k8s-xxx.yaml.

PRs ARE WELCOME

Since the environmental problems and various functionalities sometimes confuse new contributors and enthusiasts, it can be a good opportunity to check and write E2E cases especially when you are deploying the Koordinator in your local cluster for the first time.

Roadmap

• Deployment Check
  • Deploy koordlet in different versions of Kubernetes clusters.
    • Kubernetes 1.18 (minimum required version)
    • Kubernetes 1.22 ci: initialize e2e workflow #1573
    • Kubernetes 1.24 ci: initialize e2e workflow #1573
    • Kubernetes 1.26
    • Kubernetes 1.28
    • Kubernetes latest (KinD supported)
  • Deploy koordlet in different resource management modes.
    • Deploy koordlet as standalone mode. ci: support loading all components image in E2E #1705
    • Deploy koordlet as nri mode.
    • Deploy koordlet as runtime-proxy mode, also deploy koord-runtime-proxy.
• Function Test
  • Node metrics collection, estimation and reporting
    • Node Metric Reporting: verified in the "BatchResource" case. ci: add more slo controllers e2e test #1688
    • NodeTopologyReport: verified in the "CPUNormalization" case. ci: add more slo controllers e2e test #1688
    • Node Prediction: ProdReclaimable
    • CPICollector & Libpfm4
    • PSICollector
    • Accelerator
    • ColdPageCollector
    • HugePageReport
  • Runtime hook
    • BatchResource
    • CPUSetAllocator & BECPUManager
    • GroupIdentity
    • CPUNormalization
    • CoreSched
    • GPUEnvInject
  • QoS Plugin
    • RdtReconcile
    • BECPUSuppress
    • BEMemoryEvict
    • CPUBurst
    • MemoryQoS (CgroupReconcile)
    • BlkioReconcile
    • SystemConfig
• Others
  • Performance & Scalability
  • Robustness

[stale]

This issue has been automatically marked as stale because it has not had recent activity.
This bot triages issues and PRs according to the following rules:

• After 90d of inactivity, lifecycle/stale is applied
• After 30d of inactivity since lifecycle/stale was applied, the issue is closed
  You can:
• Mark this issue or PR as fresh with /remove-lifecycle stale
• Close this issue or PR with /close
  Thank you for your contributions.

[saintube]

/remove-lifecycle stale