KubeCPUOvercommit does include DaemonSets in its calculations

[dohnto]

Hi,

we have a very small cluster with just 3 nodes and the Pod CPU Request are very carefully picked so that it cluster can handle 1 node outage.

We use kubernetes-mixins, but the the alert KubeCPUOvercommit is firing for us, because it also includes DaemonSets requests in the calculations:

Example:

Node A (allocatable=2): Pod 1 (request=0.5), Daemonset 1 (request=1)
Node B (allocatable=2): Pod 2 (request=0.5), Daemonset 1 (request=1)
Node C (allocatable=2): Pod 3 (request=0.5), Daemonset 1 (request=1)

I believe, that KubeCPUOvercommit will be evaluated as follows

sum(namespace_cpu:kube_pod_container_resource_requests:sum) - (sum(kube_node_status_allocatable{resource="cpu"}) 
 - max(kube_node_status_allocatable{resource="cpu"})) > 0
and
(sum(kube_node_status_allocatable{resource="cpu"}) - max(kube_node_status_allocatable{resource="cpu"})) > 0

=>

4.5 - (6-2) > 0 and (6 - 2)

=>

0.5 > 0 and 4 > 0

But in the example provided, it should be ok to accommodate any Pod (1..3) from lost node, because the DaemonSet is not going to be scheduled.

Theoretically I could utilize the ignoringOverprovisionedWorkloadSelector, but that sounds like an misuse of this option. And ideally, we should be able to exclude the daemonsets from the calculations and ease setting up the selectors for every single daemonset, which can be very difficult.