Skip to main content

Dynamia ascend-device-plugin Chart 0.3.1

Release date: 2026-10-26

VersionPreviousCurrent
Chart0.3.00.3.1
appVersionv1.4.3v1.4.3

Chart 0.3.1 is a deployment patch for 0.3.0. It completes the configuration and mounts needed by the existing v1.4.3 application. The application image and bundled libvnpu version remain unchanged. See Chart 0.3.0 for their open-source baselines, Enterprise extensions, and application changes.

HAMi Enterprise compatibility​

Feature or modeMinimum HAMi Enterprise versionCompanion requirements and upgrade guidance
Template-based hardware partitioningv2.9.0-rc0The existing scheduler can be retained.
hami-core static soft partitioningv2.10.0-r2Retains the runtime authorization requirement introduced with application v1.4.3. This patch does not require another Enterprise upgrade when that requirement is already met.
Automatic memory expansionv2.10.0-r2Retains single-vNPU ordinary-container scope, approval and admission requirements, and the companion Chart's Kubernetes 1.30 minimum.
ENPU scheduling or hami-core compute oversell with deviceCoreScaling greater than 1v2.10.0-r5Requires the corresponding scheduler backend and resource-accounting support, with matching shared-device and node configuration. Updating the plugin Chart alone does not provide this companion functionality.

Changes​

Chart changes​

  • [Enterprise feature] Completes ENPU deployment mounts for /usr/local/enpu/vcann-rt, the per-container configuration root (default /var/lib/hami-enpu), and read-only host process information at /host/proc. Adds the corresponding enpu.* environment and configuration values, with elastic as the default ENPU policy. Optional manager settings select the node-local manager and configuration directory.
  • [Enterprise feature] Adds the explicit hamiVnpuMode backend selector and renders hamiCorePolicy, deviceCoreScaling, and enpuPolicy into bundled device configuration. An empty selector retains the legacy hamiVnpuCore.enabled fallback; a node-level mode takes precedence. Invalid values fail rendering.
  • [Enterprise feature] Changes hamiVnpuCore.coreUtilizationPolicy from force to default and adds hamiVnpuCore.defaultPolicy=spatial. Explicit force and disable retain precedence over scheduling-policy selection. This changes the default compute-control behavior even though the application version stays the same.
  • [Enterprise feature] Adds hamiVnpuCore.deviceCoreScaling, defaulting to 1, to configure advertised compute oversell. Values greater than 1 increase schedulable compute points; they do not increase physical compute capacity.

There is no application-version update in this patch. Multi-card control, the ENPU backend, and runtime authorization belong to application v1.4.3, described in the 0.3.0 release note.

Upgrade guidance​

  • Review compute-policy settings before upgrading. To retain the 0.3.0 behavior, explicitly set hamiVnpuCore.coreUtilizationPolicy=force. To adopt the new defaults, use default and spatial; check workload NPU_CORE_UTILIZATION_POLICY, Pod huawei.com/npu-sched-policy, and node hamiCorePolicy overrides. Spatial control uses cooperative core budgets and does not guarantee proportional inference throughput.
  • With config.create=false, update the HAMi-managed shared-device configuration and relevant node settings. Changing only this plugin Chart's bundled deviceConfig does not change an external ConfigMap. Use hamiVnpuMode: enpu on ENPU nodes and explicitly select huawei.com/vnpu-mode: enpu on their workloads. For hami-core, retain the matching mode and scheduling policy. Stop or reschedule affected workloads before changing a node's backend.
  • Before using ENPU or compute oversell, meet the v2.10.0-r5 companion requirement. Set deviceCoreScaling consistently in scheduler and plugin configuration, and leave it at 1 when oversell is not intended. ENPU's effective virtual-device count is capped at 100 per physical device; a container uses one physical DIE.
  • Follow the ENPU asset and workload prerequisites in the Ascend deployment guide, including compatible driver/runtime assets, the Ascend container-runtime handler, workload utilities, and node-local manager connectivity when using mem-swap. The bundled runtime provides basic soft partitioning; mem-swap needs a separately prepared supporting runtime and manager. The manager address must reach the plugin's own node.
  • Retain the default image ghcr.io/dynamia-ai/ascend-device-plugin:v1.4.3 or its matching immutable digest. If the runtime-state migration required by 0.3.0 has already been completed, this Chart-only patch does not require repeating it. Deployments still using an older application must first follow the 0.3.0 migration steps.

Apply the configuration through the existing Helm release, confirm that the DaemonSet rolls out and the plugin reloads it, then verify device registration and the selected backend before recreating workloads.

Values reference​

The parameter descriptions below are preserved from the README bundled with Chart 0.3.1. The original values.yaml from that package defines its defaults.

Expand values table
KeyTypeDefaultDescription
config.createboolfalseCreate the bundled device configuration ConfigMap for an installation without HAMi. Leave disabled when HAMi owns the device configuration.
config.deviceConfigMapNamestring"hami-scheduler-device"Name of the device configuration ConfigMap created when config.create=true.
config.existingDeviceConfigMapNamestring"hami-scheduler-device"HAMi-managed device configuration ConfigMap to reuse when config.create=false. The ConfigMap must be in the release namespace.
daemonSet.args[0]string"--config_file"
daemonSet.args[1]string"/device-config.yaml"
daemonSet.args[2]string"--v=4"
daemonSet.args[3]string"--node_config_file"
daemonSet.args[4]string"/node-config.yaml"
daemonSet.namestring"hami-ascend-device-plugin"Device plugin DaemonSet name.
deviceSplitCountint10Number of virtual devices advertised per physical Ascend NPU in hami-core or enpu mode when this chart creates the device configuration. A positive node-level vDeviceCount overrides it; ENPU caps the count at 100.
enpu.configRootstring"/var/lib/hami-enpu"Host directory for per-container ENPU configs.
enpu.defaultPolicystring"elastic"Default ENPU scheduling policy.
enpu.exposeAllDevicesboolfalseExpose all davinci devices; keep false for device isolation.
enpu.managerConfigRootstring""Manager config_dir or its vcann-rt subdirectory; empty uses configRoot.
enpu.managerURLstring""Optional node-local enpu-manager REST endpoint.
enpu.monitorPathstring"/usr/local/enpu/vcann-rt/tools/enpu-monitor"Host path to enpu-monitor.
enpu.preloadPathstring"/usr/local/enpu/vcann-rt/ld.so.preload"Host path to ld.so.preload.
enpu.runtimePathstring"/usr/local/enpu/vcann-rt/lib/libvruntime.so"Host path to libvruntime.so.
enpu.systemdDetectVirtPathstring""Optional host binary; prefer providing it in the workload image.
fullnameOverridestring""Override the fully qualified resource name.
hamiVnpuCore.coreUtilizationPolicystring"default"Default utilization override for containers without NPU_CORE_UTILIZATION_POLICY. default permits hamiCorePolicy (spatial by default); explicit force and disable retain precedence.
hamiVnpuCore.defaultPolicystring"spatial"Default hami-core compute policy: spatial, dynamic, fixed, or overdraft. Pod huawei.com/npu-sched-policy and node hamiCorePolicy override it.
hamiVnpuCore.deviceCoreScalingfloat1hami-core compute oversell ratio. The plugin advertises Devcore = round(100 * deviceCoreScaling) so HAMi can admit more than 100% of -core on one card. Values below 1 are not supported.
hamiVnpuCore.enabledboolfalseEnable hami-vnpu-core in the bundled device configuration when config.create=true. This mode requires the HAMi scheduler.
hamiVnpuModestring""Default backend: template, hami-core (alias hamiCore), or enpu. Empty uses template, or hami-core if the deprecated hamiVnpuCore.enabled is true. Node mode takes priority.
image.digeststring""Optional immutable OCI manifest digest. When set, it takes precedence over image.tag.
image.pullPolicystring"IfNotPresent"Kubernetes image pull policy.
image.repositorystring"ghcr.io/dynamia-ai/ascend-device-plugin"Container image repository.
image.tagstring""Container image tag. Defaults to the chart appVersion when empty.
nameOverridestring""Override the chart name used in resource names.
nodeConfigstring"nodes: []"Per-node hami-vnpu-core configuration written to the node ConfigMap.
nodeConfigMap.createbooltrueCreate the per-node configuration ConfigMap.
nodeConfigMap.namestring"hami-device-node-config"Per-node configuration ConfigMap name.
nodeSelector.ascendstring"on"Node label value used to schedule the device plugin.
rbac.namestring"hami-ascend"Name shared by the chart-managed RBAC resources.
resources.limits.cpustring"500m"
resources.limits.memorystring"500Mi"
resources.requests.cpustring"500m"
resources.requests.memorystring"500Mi"
runtimeClass.createbooltrueCreate the Ascend RuntimeClass.
runtimeClass.handlerstring"ascend"Container runtime handler used by the RuntimeClass.
runtimeClass.namestring"ascend"RuntimeClass resource name.
serviceAccount.createbooltrueCreate a ServiceAccount for the device plugin.
serviceAccount.namestring"hami-ascend"ServiceAccount name.
vnpuMonitor.enabledboolfalseExpose device plugin metrics and create optional Prometheus Operator resources.
vnpuMonitor.prometheusRule.createbooltrueCreate a PrometheusRule when vnpuMonitor.enabled=true.
vnpuMonitor.prometheusRule.groupLabels.vendorstring"ascend"
vnpuMonitor.prometheusRule.groupNamestring"ascend-vnpu"Recording-rule group name.
vnpuMonitor.prometheusRule.intervalstring"15s"Recording-rule evaluation interval.
vnpuMonitor.prometheusRule.labels.releasestring"prometheus"
vnpuMonitor.prometheusRule.labels.rolestring"recording-rules"
vnpuMonitor.prometheusRule.labels.vendorstring"ascend"
vnpuMonitor.prometheusRule.namestring"hami-ascend-vnpu-monitor"PrometheusRule resource name.
vnpuMonitor.prometheusRule.namespacestring"monitoring"Namespace for the PrometheusRule.
vnpuMonitor.service.namestring"hami-ascend-device-plugin-metrics"Headless Service that exposes the device plugin metrics port.
vnpuMonitor.serviceMonitor.createbooltrueCreate a ServiceMonitor when vnpuMonitor.enabled=true.
vnpuMonitor.serviceMonitor.intervalstring"15s"Prometheus scrape interval.
vnpuMonitor.serviceMonitor.labels.releasestring"prometheus"
vnpuMonitor.serviceMonitor.namestring"hami-ascend-vnpu-monitor"ServiceMonitor resource name.
vnpuMonitor.serviceMonitor.namespacestring"monitoring"Namespace for the ServiceMonitor.
vnpuMonitor.serviceMonitor.pathstring"/metrics"Metrics HTTP path.