1 of 13

SIG Node Community Meeting Update

2 of 13

What we did last cycle

  • Features
    • Alpha
      1. Ephemeral Containers support for dockershim (other runtimes handled in CRI)
    • Beta
      • Topology Manager
        1. Align CPU and devices based on node topology
        2. See https://kubernetes.io/blog/2020/04/01/kubernetes-1-18-feature-topoloy-manager-beta/
      • Startup Probes
    • GA
      • Taint Based Evictions
  • Stability / Debuggability
    • Pod Resource API fixups (active pod filtering)
    • Metric fixups (PLEG, eviction counters, etc.)
    • Pod status improvements for static pods or other edge-case scenarios

3 of 13

Plans for upcoming cycles

    • Themes
      1. Initial cgroups v2 support
      2. Iterative improvements around topology management (memory, hugepages)
      3. CRI error handling
      4. Pod termination handling improvements
      5. Kubelet cleanup
      6. RuntimeClass feedback and plans

4 of 13

How these plans affect you

    • Runtime Class
    • Scheduling Extenders for Topology Awareness

5 of 13

Things we need from you

    • Participation is always welcome!

6 of 13

kubelet

Cgroup v2 support

Topology management improvements

CRI error handling

Cleanup

Runtime Class GA requirements

7 of 13

cri-api

8 of 13

cri-tools

General improvements in testing

Additional updates pending cri-api evolution for error handling

9 of 13

node feature discovery

Testing improvements

More hardware discovery

SR-IOV

10 of 13

Related KEPs

11 of 13

Related Working Group Status

Resource Management WG is archived

12 of 13

How you can contribute

Tests over features

Tests over bug fixes

Tests with bug fixes

Test infra monitoring and health

13 of 13

Where to find us