Document and update LMA changes

[ry4nz]

SME - @nsteph

Monitoring
Monitoring – Mirantis Kubernetes Engine 4

Monitoring in both MKE 3 and MKE 4 is provided via a prometheus deployment. MKE 3 deploys prometheus as a daemonset whereas MKE 4 deploys prometheus via the kube-prometheus-stack helm chart.

The MKE 3 deployment allows placement of the prometheus servers to specific nodes with labels. MKE 4.0.0 does not offer that placement mechanism. It will be offered in a later release.

The MKE 3 docs provide instructions for using an external prometheus instance to monitor an MKE cluster. MKE 4 does not have similar instructions yet, although it should be possible to do so presently.

MKE 3 docs provide instructions for deploying grafana to the cluster. MKE 4 does not provide similar instructions, although MKE 4 does have a configuration setting that launches a grafana instance with a set of pre-configured dashboards.
Metrics Sources
MKE 3 core metrics are sourced from the ucp controller and from the docker engine. They are not provided in MKE 4.

MKE 3 component metrics and cAdvisor metrics are sourced from kubernetes and cAdvisor respectively and are provided in MKE 4. cAdvisor metrics are generally built-in to kubelet and are only provided via a standalone instance in MKE 3 to workaround limitations with the default CRI. A standalone cAdvisor is disabled by default and not necessary in MKE 4 with the current containerd CRI deployed with k0s.
Alerting
In both MKE 3 and 4, the relevant alerting component is provided through alertmanager.

In MKE 3, alertmanager is provided only as a mechanism to support opscare, specifically sending alerts to salesforce. There is no mechanism to configure alertmanager for other purposes.

In MKE4, alertmanager may be deployed using the mke 4 configuration. In MKE 4.0.0, alertmanager does not provide opscare alerting to salesforce, which will come in a later release. Support for defining additional alerting rules will also come in later versions.