Fixing a kubelet memory leak in Kubernetes 1.36

5 comments

• compumike 15 hours ago
  Author here! If you're running a Kubernetes cluster, I recommend you check `kubectl version` and see if you're running "Server Version: v1.36.[0,1,2]". If so, you may want to use the one-liner at the end of the article to check your "process_resident_memory_bytes" on each node, and consider restarting kubelet as a temporary workaround to tame the memory leak until v1.36.3 is released.
• __turbobrew__ 21 minutes ago
  A good reason to health check the kubelet process and restart it when the checks fail.
  [-]
  • compumike 5 minutes ago
    What kind of health checks? In my case, the kubelet process was staying alive and responsive to queries, I believe due to:

      # cat /proc/$(pgrep kubelet)/oom_score_adj
      -999
      
      (from OOMScoreAdjust=-999 in /etc/systemd/system/kubelet.service)  
    

    With this score, the Linux OOM killer wouldn't touch it, but any of my Pods were fair game.
• rirze 1 hour ago
  Very cool. It's often daunting to contribute to such a well-established and recognizable project, but this is exactly how it should work.
• CamouflagedKiwi 59 minutes ago
  Nice find.

  Can't help but feel this is one of the subtle traps hidden beneath the advice that contexts aren't supposed to be stored. I know it's not always that easy, of course.

  [-]
  • compumike 30 minutes ago
    Thanks. I know there's a `go vet` tool that's run as part of Kubernetes CI, and one of its checks is:

      lostcancel: check cancel func returned by context.WithCancel is called
    

    I'm not 100% sure why `go vet` didn't catch this issue, but storing the cancelFn in the struct is probably part of the reason. Any Go experts know if that's the case?
• fsuts 36 minutes ago
  Not all heroes wear capes! Well done