This new version makes significant changes to the number of components that are deployed to the cluster, and introduces many new configuration options to tune the behavior to your environment. We encourage you to take a look at what's changed in full detail here.
The number and format of the metrics reported by version 3 of the integration have not changed with respect to earlier versions.
- The format of the
values.ymlfile has changed to accommodate the newly added configuration options. Please take a look at our migration guide to see how to change your configuration.
- Our solution is now deployed in three components:
DaemonSetto monitor the Kubelet, deployed in all nodes of the cluster.
- A second
DaemonSetto monitor the control plane, deployed in master nodes only.
Deploymentto collect metrics from kube-state-metrics, deployed in the same node as the latter.
- We now offer better control for CPU and memory limits and requests, which can be now configured for the three components individually.
- Impact of discovery and collection operations on the API server has been greatly reduced, thanks to the use of kubernetes informers.
- Logs messages have been greatly revamped to surface problems more clearly.
- Comprehensive configuration options have been added to provide fine-grain control to how the integration discovers and connects to metric providers. Remarkably:
- Discovery options for control plane components have been improved. You can check the details on how discovery is configured here.
- It is now possible to collect metrics from control plane components running outside of the cluster.
- Discovery options for KSM and the kubelet have also been added.
- The interval at which metrics are collected is now configurable.