Effective Monday, 1 January 2022, our Kubernetes integration drops support for Kubernetes v1.15 and lower. The Kubernetes integration v2.8.0 and higher will only be compatible with Kubernetes versions 1.16 and higher. For more information, read this note or contact your account team.
Background
Enabling compatibility with the latest Kubernetes versions and adding new features to our Kubernetes offering prevents us from offering first-class support to versions prior to v1.16.
What is happening?
- The latest Kubernetes version v1.22 has API incompatibilities with versions prior to v1.16.
- Most major Kubernetes cloud providers have already deprecated v1.15 and lower.
What do you need to do?
It's easy: Upgrade your Kubernetes clusters to a supported version.
What happens if you don't make any changes to your account?
The Kubernetes integration may continue to work with end-of-lifed versions. However, we can't guarantee the quality of the solution as new releases may cause some incompatibilities.
Note that support requests regarding these end-of-lifed versions won't be accepted.
Changed
- Updated agent and integrations to their latest version
Changed
- Updated agent and integrations to their latest versions
Breaking changes
- With the addition of Kubernetes 1.22.x support, we are phasing out of the support bracket Kubernetes versions 1.15.x and below. This means the supported versions bracket for this version and the rest of the nri-kubernetes 2.8.x branch is from Kubernetes v1.16.x to v1.22.x.
- Users may still use previous versions of the integration and their Helm charts to instrument their clusters, and limited support will be offered for them. Development for said older versions is considered stopped and New Relic does not guarantee feature parity or bugfixes (security or otherwise) for those older versions.
Changed
- Node status and conditions are now fetched from the API Server rather than KSM, which fixes some inconsistencies in the samples. This does not change which data is reported, and should be an invisible change. (#194).
- Add a series of parameters which allow to configure a jitter to be applied to API Server response caching, which might help to spread the load on large clusters. (#185).
Fixed
- Node status and conditions are now fetched from the Kubernetes API rather than KSM, which makes the query experience more consistent.
- Infrastructure agent and integrations have been updated to the latest versions.
Added
- Integration now reports node status and conditions as
condition.{Name}
(for examplecondition.Ready
,condition.PIDPressure
). - Added the new
KubeStateMetricsNamespace
parameter to restrict discovery of KSM pod to a particular namespace.- This should help reduce load in the control plane for clusters with many pods and/or nodes.
- Added the
unschedulable
metric to K8sNodeSamples. - Added the
podIP
metric to K8sPodSample.
Changed
- Upgrade
infrastructure-bundle
to 2.6.0 #123- See https://github.com/newrelic/infrastructure-bundle/releases/tag/2.6.0 for more details about the upgraded integrations in this release of the
infrastructure-bundle
- See https://github.com/newrelic/infrastructure-bundle/releases/tag/2.6.0 for more details about the upgraded integrations in this release of the
Changelog
Added
- Added support for ARM and ARM64 nodes, and published a multiarch docker image.
Changelog
Fixed
- Correctly identifing the Kubernetes server version with characters (#81).