Steps to reproduce:

Apply "Kubernetes cluster state by HTTP" template to a Kubernetes host and successfully configure it.

Result:
Observe Problems being generated whenever the desired number of replicas for an object are not met, however brief the condition might be.

Expected:
There should be macros that allow the thresholds for problem generation on replica mismatches to be adjusted. The problem formulas currently use a hardcoded logic of last(X)-last(Y)<>0, which uses the most recently stored value for items X and Y and does not allow for a custom threshold.

The current logic can generate a lot of unnecessary Problems during normal operation of the cluster. If the reduced number of ready replicas is a result of a restart or redeployment action (particularly for deployments that use a rollout strategy of "Recreate"), the managed replicaSets will generate monitoring noise for a short-lived condition that quickly goes away.

The following autodiscovery triggers should be re-evaluated:

Kubernetes: Namespace {#NAMESPACE} RS {#NAME}: ReplicasSet mismatch
Kubernetes: Namespace {#NAMESPACE} RS {#NAME}: Statefulset replicas mismatch
Kubernetes: Namespace {#NAMESPACE} Deployment {#NAME}: Deployment replicas mismatch