v0.4.3

• Update calculation for GB in MIG profile name
• Make the systemd-mig-manager a dependency of systemd-resolved.service

v0.4.2

• Update CUDA image to 11.7.0
• Add extra assert in k8s-mig-manager to double check mig-mode change applied
• Update mig-manager image to use NGC DL license

v0.4.1

• Keep NVML alive across all mig-parted commands (except GPU reset)
• Remove unnecessary services from hooks.sh

v0.4.0

• Update nvidia-mig-parted.sh to include MIG_PARTED_CHECKPOINT_FILE
• Add checkpoint / restore commands to mig-parted CLI
• Update golang version to 1.16.4
• Support instantiation of *_PROFILE_6_SLICE GIs and CIs
• Update cyrus-sasl-lib to address CVE-2022-24407
• Add support for MIG profiles with +me as an attribute extension
• Support Compute Instances in mig-parted config such that CI != GI
• Update go-nvml to v0.11.6
• Change semantics of all to mean all-mig-capable in mig-parted config

v0.3.0

• k8s-mig-manager: Add support for multi-arch images
• k8s-mig-manager: Handle eviction of NVSM pod when applying MIG changes

v0.2.0

• nvidia-mig-parted: Support passing newer GI and CI profile enums on older drivers
• k8s-mig-manager: Rename nvcr.io/nvidia to nvcr.io/nvidia/cloud-native
• k8s-mig-manager: Add support for pre-installed drivers
• systemd-mig-manager: Update logic to remove 'containerd' containers in utils.sh
• systemd-mig-manager: Update logic to shutdown only active systemd services in list
• ci-infrastructure: Rework build and CI to align with other projects
• ci-infrastructure: Use pulse instead of contamer for scans

v0.1.3

• Update to latest cuda base images
• Honor existing deployment state of GPU clients in K8s when reconfiguring
• Add license file to reeconfigure-mig.sh
• Add default configs for the PG506-96GB card
• Remove CombinedMigManager and add wrappers for Mode/Config Managers
• Add a function to check the minimum NVML version required
• Add SystemGetNVMLVersion() to the NVML interface
• Fix small bug in assert logic for non MIG-capable GPUs

v0.1.2

• Do not start nvidia-mig-manager.service when installing the .deb
• Restore lost assert_gpu_reset_available() function
• Add nvidia-dcgm.service to driver_services array
• Split dcgm, and dcgm-exporter in k8s-mig-manager

v0.1.1

• Pass host-root-mount to k8s-mig-manager to properly support reboot of node for GPU passthrough virtualization
• Update config.yaml for systemd-mig-manager with support for more devices
• Delete validator pod only after successfully applying MIG config in k8s-mig-manager
• Always reenable operator components torn down by k8s-mig-manager, even on failure

v0.1.0

Initial release of nvidia-mig-parted.

This release also includes:

• A release of the k8s-mig-manager for managing MIG through the Kubernetes GPU Operator.
  https://hub.docker.com/repository/docker/nvidia/k8s-mig-manager
  https://ngc.nvidia.com/containers/nvidia:cloud-native:k8s-mig-manager

• A release of the systemd-based mig-manager for bare-metal management of MIG.
  Debian, RPM, and tarball packages for this are included in the assets below