What's Changed

• Ensure that cdi.FeatureFlags are passed to CDI library
• Fix race condition in config-manager when label is unset
• Fix nested container use cases by ensuring that IPC sockets are not mounted readonly
• Bump NVIDIA Container Toolkit to v1.18.2
• Bump distroless base image to v3.2.2-dev

Full Changelog: v0.18.1...v0.18.2

Changelog

• Allow CDI feature flags to be set
• Pass driver root to nvinfo.New in device plugin main
• Bump NVIDIA Container Toolkit to v1.18.1
• Bump distroless base image to v3.2.1-dev
• Bump github.com/opencontainers/selinux from 1.12.0 to 1.13.1 (#1506)

Full Changelog: v0.18.0...v0.18.1

Changelog

• Rename getHealthCheckXids and clarify documentation
• Add support for explicitly enabling XIDs in health checks
• Deduplicate requested device IDs
• Check for nil before reading boolean config values
• Make gated modes (GDS, MOFED, GDRCOPY) optional in CDI
• Add support for setting gdrcopyEnabled
• Ignore errors getting device memory using NVML
• Ensure that directory volumes have Directory type
• Switch to plain golang image for builds
• Remove unneeded intermediate container
• Update CI definitions
• Switch to distroless golang image
• Update README.md with RuntimeClass
• Pass a single context throughout the device-plugin method call stack (#1284)
• Remove internal logger in favour of klog (#1277)
• Remove FAIL_ON_INIT_ERROR from static examples
• Detect blackwell architecture
• Updated .release:staging to stage device-plugin images in nvstaging
• Use MiB instead of MB for gpu-memory
• Ignore XID error 109
• Update README.md adjust set docker runtime default
• Remove nvidia.com/gpu.imex-domain label
• Fix containerd runc config error when creating a kind cluster
• Use stable nividia-container-toolkit repo when creating a kind cluster
• Switch to context package in go stdlib
• Raise a warning instead of an error if GPU mode labeler fails
• Add ada-lovelace architecture label for compute capability 8.9
• Ensure FAIL_ON_INIT_ERROR boolean env is quoted
• Honor fail-on-init-error when no resources are found
• Enable hostPID in the mps-control-daemon pod (#1045)

Full Changelog: v0.17.1...v0.18.0

What's Changed

• Bump slackapi/slack-github-action from 2.1.0 to 2.1.1 by @dependabot[bot] in #1317
• Bump github.com/NVIDIA/go-nvlib from 0.7.2 to 0.7.4 by @dependabot[bot] in #1346
• Bump golang from 1.23.11 to 1.23.12 in /deployments/devel by @dependabot[bot] in #1355
• Ensure that directory volumes have Directory type by @elezar in #1368
• Bump nvidia/cuda from 12.9.1-base-ubi9 to 13.0.0-base-ubi9 in /deployments/container by @dependabot[bot] in #1369
• Ignore errors getting device memory using NVML by @elezar in #1374
• Bump project version to v0.17.4 by @cdesiniotis in #1402
• [no-relnote] update ngc publishing logic for release pipelines by @cdesiniotis in #1406

Full Changelog: v0.17.3...v0.17.4

What's Changed

• Bump github.com/NVIDIA/nvidia-container-toolkit from 1.17.6 to 1.17.8 by @dependabot[bot] in #1275
• Bump nvidia/cuda from 12.9.0-base-ubi9 to 12.9.1-base-ubi9 in /deployments/container by @dependabot[bot] in #1300
• Bump github.com/NVIDIA/go-nvml from 0.12.4-1 to 0.12.9-0 by @dependabot[bot] in #1287
• Bump golang from 1.23.9 to 1.23.10 in /deployments/devel by @dependabot[bot] in #1283
• Bump golang from 1.23.10 to 1.23.11 in /deployments/devel by @dependabot[bot] in #1318
• Bump release v0.17.3 by @elezar in #1326
• Backport: Bump golang.org/x/oauth2 from 0.23.0 to 0.27.0 by @cdesiniotis in #1328
• Updated .release:staging to stage device-plugin images in nvstaging by @elezar in #1329

Full Changelog: v0.17.2...v0.17.3

What's Changed

• Update nvidia.com/gpu.product label to include blackwell architectures
• Update documentation to indicate that nvidia.com/gpu.memory label is in MiB instead of MB

Full Changelog: v0.17.1...v0.17.2

What's Changed

• Bump golang from 1.23.2 to 1.23.3 in /deployments/devel by @dependabot in #1063
• Bump the k8sio group across 1 directory with 5 updates by @dependabot in #1066
• Ensure FAIL_ON_INIT_ERROR boolean env is quoted by @elezar in #1076
• Bump nvidia/cuda from 12.6.2-base-ubi9 to 12.6.3-base-ubi9 in /deployments/container by @dependabot in #1084
• Bump github.com/NVIDIA/nvidia-container-toolkit from 1.17.0 to 1.17.2 by @dependabot in #1068
• Bump google.golang.org/grpc from 1.65.0 to 1.65.1 by @dependabot in #1069
• Bump sigs.k8s.io/node-feature-discovery from 0.15.4 to 0.15.7 by @dependabot in #1070
• Bump NVIDIA/holodeck from 0.2.3 to 0.2.4 by @dependabot in #1064
• Honor fail-on-init-error when no resources are found by @elezar in #1061
• Bump github.com/opencontainers/selinux from 1.11.0 to 1.11.1 by @dependabot in #1067
• Add ada-lovelace architecture label for compute capability 8.9 by @elezar in #1090
• Switch to context package in go stdlib by @elezar in #1114
• Bump github.com/NVIDIA/nvidia-container-toolkit from 1.17.2 to 1.17.4 by @dependabot in #1138
• Bump nvidia/cuda from 12.6.3-base-ubi9 to 12.8.0-base-ubi9 in /deployments/container by @dependabot in #1142
• Bump NVIDIA/holodeck from 0.2.4 to 0.2.5 by @dependabot in #1131
• Bump slackapi/slack-github-action from 1.27.0 to 2.0.0 by @dependabot in #1065
• Bump github.com/NVIDIA/go-nvlib from 0.7.0 to 0.7.1 by @dependabot in #1151
• Ignore XID error 109 by @elezar in #1171
• Remove nvidia.com/gpu.imex-domain label by @elezar in #1152
• Bump azure/setup-helm from 4.2.0 to 4.3.0 by @dependabot in #1176
• Bump github.com/NVIDIA/nvidia-container-toolkit from 1.17.4 to 1.17.5-rc.1 by @elezar in #1192

Full Changelog: v0.17.0...v0.17.1

What's Changed

v0.17.0

• Promote v0.17.0-rc.1 to GA

v0.17.0-rc.1

• Add CAP_SYS_ADMIN if volume-mounts list strategy is included
• Remove unneeded DEVICE_PLUGIN_MODE envvar
• Fix applying SELinux label for MPS
• Use a base image that aligns with the ubi-minimal base image
• Switch to a ubi9-based base image
• Remove namespace field from cluster-scoped resources
• Generate labels for IMEX cligue and domain
• Add optional injection of the default IMEX channel
• Allow kubelet-socket to be specified as command line argument

What's Changed

• Add CAP_SYS_ADMIN if volume-mounts list strategy is included
• Remove unneeded DEVICE_PLUGIN_MODE envvar
• Fix applying SELinux label for MPS
• Use a base image that aligns with the ubi-minimal base image
• Switch to a ubi9-based base image
• Remove namespace field from cluster-scoped resources
• Generate labels for IMEX cligue and domain
• Add optional injection of the default IMEX channel
• Allow kubelet-socket to be specified as command line argument

What's Changed

• Fix applying SELinux label for MPS
• Remove unneeded DEVICE_PLUGIN_MODE envvar
• Add CAP_SYS_ADMIN if volume-mounts list strategy is included (fixes #856)

Full Changelog: v0.16.1...v0.16.2