Maximum number of pods per node

pup_seba · April 7, 2021, 7:56am

Hi,

Kubernetes web page here mentions that the “maximum amount of pods is x”, but it does not mention if we are talking about running pods or “leftover” like the ones that remain after a job is exectued and the pod is marked as completed.

So, is this reccomendation for “total pods” or only for “running pods”?

tej-singh-rana · April 8, 2021, 5:26pm

Hey @pup_seba
I think it’s about the total number of pods (running, having an issue in the pod, completed and something like that), which consumes space in the node.

pup_seba · April 22, 2021, 6:49am

Hi @tej-singh-rana,

If that were the case, then it would depend on the characteristics (disk space available for local data, ram, etc) of such host and saying “100” would be way too arbitrary imho.

We are facing some issues with kubelet PLEG component and its gRPC messages exceeding the 16MB limit when too many container info is in the node, since the message grows too large and PLEG fails to report to kubelet making the node NotReady for kubernetes.

Keeping a low number of container elements at the host level helps aliviate the situation. I wonder if this recomendation is related to this situation, or why is even there and why it is a “100” and not “200” or “50”.

kong_chen · October 19, 2022, 5:58am

Please take a look at this comment, it depends on pod status

github.com

kubernetes/kubernetes/blob/master/pkg/quota/v1/evaluator/core/pods.go#L331


      
          	switch selector.ScopeName {
          	case corev1.ResourceQuotaScopeTerminating:
          		return isTerminating(pod), nil
          	case corev1.ResourceQuotaScopeNotTerminating:
          		return !isTerminating(pod), nil
          	case corev1.ResourceQuotaScopeBestEffort:
          		return isBestEffort(pod), nil
          	case corev1.ResourceQuotaScopeNotBestEffort:
          		return !isBestEffort(pod), nil
          	case corev1.ResourceQuotaScopePriorityClass:
          		if selector.Operator == corev1.ScopeSelectorOpExists {
          			// This is just checking for existence of a priorityClass on the pod,
          			// no need to take the overhead of selector parsing/evaluation.
          			return len(pod.Spec.PriorityClassName) != 0, nil
          		}
          		return podMatchesSelector(pod, selector)
          	case corev1.ResourceQuotaScopeCrossNamespacePodAffinity:
          		return usesCrossNamespacePodAffinity(pod), nil
          	}
          	return false, nil
          }

kong_chen · October 19, 2022, 7:26am

Terminal pods are excluded in scheduler

Topic		Replies	Views
Max pod calculation per node General Discussions	0	398	May 17, 2023
Imbalance of pods in Kubernetes General Discussions	1	1080	October 27, 2019
GKE - More than enough host resource capacity available and pods becoming unschedulable General Discussions	5	2725	April 8, 2019
We have exceeded the limit on the number of pods and need a way to enable kubernetes to support running more containers General Discussions	5	917	June 17, 2022
Why a limit of 110 pods by worker? General Discussions	1	811	May 16, 2019

Maximum number of pods per node

Related topics