In correlation with the end of life (EOL) for MKE 3.7.x, maintenance of this documentation set was discontinued as of 2025-AUG-29. Click here for the latest MKE 3.x version documentation.

Addressed issues

Issues addressed in the MKE 3.7.12 release include:

  • [FIELD-7110] Fixed an issue wherein nvidia-gpu-feature-discovery pods crashed with the "--mig-strategy=mixed": executable file not found in $PATH: unknown error whenever nvidia_device_plugin was enabled. If you observe related issues, apply the workaround described in [MKE-11535] ucp-nvidia-gpu-feature-discovery pods may enter CrashLoopBackOff state.

  • [FIELD-7106] At startup, the ucp-kubelet container now obtains the profile label directly from etcd, thus ensuring that the kubelet profile settings are in place at the moment the container is created.

  • [FIELD-7059] Fixed an issue wherein the ucp-cluster-agent container leaks connections to Windows nodes.

  • [FIELD-7053] Fixed an issue wherein, in rare cases, cri-dockerd failed to pull large images.

  • [FIELD-7037] Fixed an issue exclusive to MKE 3.7.8 through 3.7.10, wherein in an air-gapped environment, the addition of a second manager node could cause cluster deployment to fail.

  • [FIELD-7032] Fixed an issue wherein SAML metadata was not deleted from RethinkDB upon disablement of the SAML configuration.

  • [FIELD-7012] Addition of TLS configuration to node-exporter.