A DaemonSet should create a pod on each node. You can limit it further to only match specific hosts, but I don't think that can be implemented in kuota-calc without providing all node info, which seems excessive. But it would be reasonable to provice the node count and just assume that a pod will be created on all nodes.
Also DaemonSets have an UpdateStrategy, which defaults to RollingUpdate with MaxUnavailable 1 and MaxSurge 2. This means it should immediately shut down the old pod and start the new one. This means, for a short time, two pods per node are running -> the current implementation only assumes one pod, for all nodes.
You don't actually need the quotas for 2 pods, as it then just would wait for the old one to shutdown before starting the new one, so one could argue that the current implementation is fine. But at least the node count needs consideration.
A DaemonSet should create a pod on each node. You can limit it further to only match specific hosts, but I don't think that can be implemented in kuota-calc without providing all node info, which seems excessive. But it would be reasonable to provice the node count and just assume that a pod will be created on all nodes.
Also DaemonSets have an UpdateStrategy, which defaults to RollingUpdate with MaxUnavailable 1 and MaxSurge 2. This means it should immediately shut down the old pod and start the new one. This means, for a short time, two pods per node are running -> the current implementation only assumes one pod, for all nodes.
You don't actually need the quotas for 2 pods, as it then just would wait for the old one to shutdown before starting the new one, so one could argue that the current implementation is fine. But at least the node count needs consideration.