HAMi Enterprise 2.7.1-rc0

Chart version	Declared appVersion	Release date
2.7.1-rc0	2.7.1-rc0	2026-01-16

This is Dynamia's first HAMi Enterprise release and is a pre-release. This page uses open-source HAMi v2.7.0 as its baseline.

Software changes

Enterprise extensions compared with open-source HAMi v2.7.0

Pass NVIDIA device-allocation results through the Downward API. The scheduler writes GPU UUIDs, memory limits, and compute limits to Pod annotations and then passes them to the corresponding containers through the Kubernetes Downward API. The regular NVIDIA vGPU allocation path can therefore skip the step that relies on a node lock to find the Pod awaiting allocation. nvidia.useDownward is true in the default device configuration. The companion device plugin, Enterprise runtime library, and NVIDIA container runtime environment are still required.

Adjust the memory limit of a running container. The Enterprise monitor receives memory-adjustment requests reported by the runtime and submits them to the scheduler, which checks available device capacity. After the scheduler updates allocation records and quota usage, the monitor synchronizes the approved limit back to the runtime. This capability requires the companion Enterprise runtime, monitor, and scheduler. It does not modify the Pod's resources.limits, and it does not mean all workloads will expand automatically. The first release does not additionally check the namespace ResourceQuota limit when deciding whether to expand memory; do not rely on it as a hard quota boundary for runtime expansion.

Manage NVIDIA device licenses. The scheduler includes device-license validity in device selection, and the monitor synchronizes license updates. The scheduler metric vGPULicenseRemains reports the remaining license time in seconds and can be used to configure expiration alerts. Obtain license materials for installation and renewal through the agreed delivery channel.

Expose container memory details. In addition to total memory usage and limits, Enterprise monitoring provides the following metrics in bytes. These are current values and should be treated as Gauges:

Metric	Meaning
hami_container_device_memory_context_bytes	Context memory usage
hami_container_device_memory_data_bytes	Data memory usage
hami_container_device_memory_module_bytes	Module memory usage
hami_container_device_memory_offset_bytes	Offset memory usage

These metrics associate containers and devices using the podnamespace, podname, ctrname, vdeviceid, and deviceuuid labels. Use these exact label names in collection rules.

Upstream changes included in this release

This release also includes the following community changes made after open-source HAMi v2.7.0.

Change	Impact on usage and integration	Upstream PR
Multi-model configuration for Iluvatar devices	Configure device count, memory, and compute resource names separately for BI-V100, BI-V150, MR-V100, and MR-V50. Requires the matching GPU Manager and devices.iluvatar.enabled: true (default false). For example, BI-V150 uses iluvatar.ai/BI-V150-vgpu, iluvatar.ai/BI-V150.vMem, and iluvatar.ai/BI-V150.vCore.	HAMi #1399
Online/offline task classification for Metax sGPU	The Pod annotation metax-tech.com/sgpu-app-class accepts online or offline to distinguish device selection for these task types. When this annotation is set, each applicable container can request only one sGPU. Task execution priority still depends on the companion Metax components.	HAMi #1436
AMD MI300X full-card resources	Requires a separate compatible AMD device plugin. HAMi uses the reported amd.com/gpu capacity to include full-card allocations in scheduler records and metrics. Memory accounting is fixed at 192000 MiB per card. These metrics represent scheduler allocation amounts, not real-time memory usage or utilization telemetry. This release does not provide AMD memory or compute partitioning.	HAMi #1472
Mixed-size selection for dynamic MIG	When a partition profile includes multiple instance sizes, matches each request to a suitable instance, fixing cases where a small-memory request incorrectly occupied a larger instance.	HAMi #1518
Quota usage cleanup	Reclaims quota usage when a Pod enters a terminating state. When a ResourceQuota is updated, clears the old configuration first so removed quota entries do not continue to impose stale limits.	HAMi #1400
Track GPU counts per node	Tracks NVIDIA GPU counts reported by each node separately, preventing nodes from overwriting each other's device-count records.	HAMi #1555
Node-lock annotation parsing	Returns a clear error when a node-lock annotation contains a delimiter but has an invalid number of fields; the legacy timestamp-only format remains parseable. The Enterprise default Downward API path does not depend on this node-lock allocation flow.	HAMi #1560
Node configuration and selector fixes	Writes node configuration to the ConfigMap as JSON text and consistently uses devicePlugin.nvidiaNodeSelector as the NVIDIA node selector. Check key names when migrating old values.	HAMi #1558, HAMi #1546
Optional mock device plugin	Installs a mock device plugin for scheduling validation without a GPU. It is disabled by default; mock devices do not perform real GPU computation.	HAMi #1534
Optional DRA installation path	The main Chart includes the hami-dra subchart 0.0.1, which can switch deployment to the Kubernetes Dynamic Resource Allocation path. It is disabled by default. See below for the behavior when switching paths.	HAMi #1542

Chart changes

The first main Chart is named hami; the default application image is ghcr.io/dynamia-ai/hami:v2.7.1-rc0. This tag is used for the scheduler extender, NVIDIA device plugin, and vGPU monitor. The bundled kube-scheduler, certificate helper, and optional subcharts use their own image settings.

Setting	Default	Description
schedulerName	hami-scheduler	Scheduler name used by workloads
scheduler.replicas	1	Deploy one scheduler Pod by default
devicePlugin.nvidiaNodeSelector	gpu: "on"	Deploy the NVIDIA device plugin to nodes with this label
devicePlugin.deviceSplitCount	10	Default number of device-sharing slices
devicePlugin.deviceMemoryScaling	1	Do not scale up the default allocatable memory capacity
devicePlugin.deviceListStrategy	envvar	Pass the device list through environment variables by default
mockDevicePlugin.enabled	false	When enabled, use mock plugin image tag 0.1.0 by default
dra.enabled	false	When enabled, use the DRA subchart instead of the regular scheduler and NVIDIA device plugin from the main Chart

The DRA path must meet the requirements of the companion driver and Kubernetes DRA. Enterprise extension notes for the regular device-plugin path do not automatically apply to it.

Upgrade notes

Install the first release or migrate from open source

Use the delivered hami-2.7.1-rc0.tgz, the companion Enterprise runtime, and license materials. If you use a custom image registry or explicit image tags, check the versions of the scheduler extender, device plugin, and monitor together; do not replace only one component.

When migrating from an open-source deployment, check node selection, scheduler integration, and device resource requests against this release's configuration. If old values use devicePlugin.nvidianodeSelector, change the key to devicePlugin.nvidiaNodeSelector. The default selector is gpu: "on"; omitting a custom selector may deploy the plugin to the wrong set of nodes.

The old Iluvatar settings iluvatarResourceName, iluvatarResourceMem, and iluvatarResourceCore are no longer used by the Chart, and old values are not migrated automatically. The former default resources iluvatar.ai/vgpu, iluvatar.ai/vcuda-memory, and iluvatar.ai/vcuda-core are replaced with model-specific resource names. Set devices.iluvatar.enabled: true, update Pod device-count, memory, and compute requests for the actual model, and deploy the companion GPU Manager with the same resource names. Replacing only the Chart while keeping old Pod resource names does not complete the migration.

devicePlugin.nodeConfiguration.config accepts JSON text. Set name to the actual node name; the default example your-node-name does not automatically match a node in your cluster. When devicePlugin.nodeConfiguration.externalConfigName is set, the Chart uses the specified ConfigMap and does not create this node configuration.

Check the Downward API and startup hook

The default device configuration sets both nvidia.useDownward: true and nvidia.postNvidiaSmi: true. With the Downward API enabled, the latter adds a postStart hook that runs nvidia-smi in containers requesting NVIDIA devices. The application image must provide this command. An existing postStart hook is replaced, not merged automatically.

To disable the startup hook, keep nvidia.useDownward: true in the device configuration and set nvidia.postNvidiaSmi to false. These two fields are not top-level Helm parameters in this release. If you use a custom files/device-config.yaml, it replaces the entire default device configuration; preserve the other device fields your deployment uses rather than supplying only these two. Update the scheduler and device plugin with the modified configuration, then recreate workload Pods whose injection behavior must change.

The first release handles NVIDIA allocation annotations using the regular-container list and does not yet include the later protections for empty allocation lists and out-of-range container indexes. Combinations involving non-GPU containers, init containers, or native sidecars should not be considered fully supported on this basis.

Check quotas and monitoring integrations

The first release does not include the NVIDIA multi-container quota-aggregation fix. Multiple GPU containers in one Pod may each pass their individual checks even when their combined request exceeds the namespace's remaining quota. Deployments that need this fix can refer to Chart 2.8.0-rc0. The quota boundary for runtime memory adjustment is described above.

Memory-detail metrics come from the vGPU monitor; vGPULicenseRemains comes from the scheduler. Configure collection for both components and use current-value semantics in dashboards and alerts. This release still retains older metrics such as HostGPUMemoryUsage and HostCoreUtilization; do not remove their collection rules based on a removal list from a later release.

See the Chart and appVersion mapping or the software center.

Values reference

The following parameter descriptions are taken from the README bundled with the Chart 2.7.1-rc0 release package and are preserved as published. Use this version's values.yaml as the source of default configuration.

Expand values table

Global Configuration

Parameter	Description	Default Value
global.imageRegistry	Global Docker image registry	""
global.imagePullSecrets	Global Docker image pull secrets	[]
global.imageTag	Image tag	"v2.7.1"
global.gpuHookPath	GPU Hook path	/usr/local
global.labels	Global labels	{}
global.annotations	Global annotations	{}
global.managedNodeSelectorEnable	Whether to enable managed node selector	false
global.managedNodeSelector.usage	Managed node selector usage	"gpu"
nameOverride	Name override	""
fullnameOverride	Full name override	""
namespaceOverride	Namespace override	""

Resource Name Configuration

NVIDIA GPU Resources

Parameter	Description	Default Value
resourceName	GPU resource name	"nvidia.com/gpu"
resourceMem	GPU memory resource name	"nvidia.com/gpumem"
resourceMemPercentage	GPU memory percentage resource name	"nvidia.com/gpumem-percentage"
resourceCores	GPU core resource name	"nvidia.com/gpucores"
resourcePriority	GPU priority resource name	"nvidia.com/priority"

Cambricon MLU Resources

Parameter	Description	Default Value
mluResourceName	MLU resource name	"cambricon.com/vmlu"
mluResourceMem	MLU memory resource name	"cambricon.com/mlu.smlu.vmemory"
mluResourceCores	MLU core resource name	"cambricon.com/mlu.smlu.vcore"

Hygon DCU Resources

Parameter	Description	Default Value
dcuResourceName	DCU resource name	"hygon.com/dcunum"
dcuResourceMem	DCU memory resource name	"hygon.com/dcumem"
dcuResourceCores	DCU core resource name	"hygon.com/dcucores"

Metax GPU Resources

Parameter	Description	Default Value
metaxResourceName	GPU resource name	"metax-tech.com/sgpu"
metaxResourceCore	GPU core resource name	"metax-tech.com/vcore"
metaxResourceMem	GPU memory resource name	"metax-tech.com/vmemory"
metaxsGPUTopologyAware	GPU topology awareness	"false"

Enflame GCU Resources

Parameter	Description	Default Value
enflameResourceNameVGCU	vGCU resource name	"enflame.com/vgcu"
enflameResourceNameVGCUPercentage	vGCU percentage resource name	"enflame.com/vgcu-percentage"

Kunlunxin XPU Resources

Parameter	Description	Default Value
kunlunResourceName	XPU resource name	"kunlunxin.com/xpu"

Scheduler Configuration

Parameter	Description	Default Value
schedulerName	Scheduler name	"hami-scheduler"
scheduler.nodeName	Define node name, scheduler will schedule to this node	""
scheduler.overwriteEnv	Whether to overwrite environment variables	"false"
scheduler.defaultSchedulerPolicy.nodeSchedulerPolicy	Node scheduler policy	binpack
scheduler.defaultSchedulerPolicy.gpuSchedulerPolicy	GPU scheduler policy	spread
scheduler.metricsBindAddress	Metrics bind address	":9395"
scheduler.forceOverwriteDefaultScheduler	Whether to force overwrite default scheduler	true
scheduler.livenessProbe	Whether to enable liveness probe	false
scheduler.leaderElect	Whether to enable leader election	true
scheduler.replicas	Number of replicas	1

Kube Scheduler Configuration

Parameter	Description	Default Value
scheduler.kubeScheduler.enabled	Whether to run kube-scheduler container in scheduler pod	true
scheduler.kubeScheduler.image.registry	Kube scheduler image registry	"registry.cn-hangzhou.aliyuncs.com"
scheduler.kubeScheduler.image.repository	Kube scheduler image repository	"google_containers/kube-scheduler"
scheduler.kubeScheduler.image.tag	Kube scheduler image tag	""
scheduler.kubeScheduler.image.pullPolicy	Kube scheduler image pull policy	IfNotPresent
scheduler.kubeScheduler.image.pullSecrets	Kube scheduler image pull secrets	[]
scheduler.kubeScheduler.extraNewArgs	Extra new arguments	["--config=/config/config.yaml", "-v=4"]
scheduler.kubeScheduler.extraArgs	Extra arguments	["--policy-config-file=/config/config.json", "-v=4"]

Extender Configuration

Parameter	Description	Default Value
scheduler.extender.image.registry	Scheduler extender image registry	"docker.io"
scheduler.extender.image.repository	Scheduler extender image repository	"projecthami/hami"
scheduler.extender.image.tag	Scheduler extender image tag	""
scheduler.extender.image.pullPolicy	Scheduler extender image pull policy	IfNotPresent
scheduler.extender.image.pullSecrets	Scheduler extender image pull secrets	[]
scheduler.extender.extraArgs	Scheduler extender extra arguments	["--debug", "-v=4"]

Admission Webhook Configuration

Parameter	Description	Default Value
scheduler.admissionWebhook.enabled	Whether to enable admission webhook	true
scheduler.admissionWebhook.customURL.enabled	Whether to enable custom URL	false
scheduler.admissionWebhook.customURL.host	Custom URL host	127.0.0.1
scheduler.admissionWebhook.customURL.port	Custom URL port	31998
scheduler.admissionWebhook.customURL.path	Custom URL path	/webhook
scheduler.admissionWebhook.reinvocationPolicy	Reinvocation policy	Never
scheduler.admissionWebhook.failurePolicy	Failure policy	Ignore

TLS Certificate Configuration

Parameter	Description	Default Value
scheduler.certManager.enabled	Whether to use cert-manager to generate self-signed certificates	false
scheduler.patch.enabled	Whether to use kube-webhook-certgen to generate self-signed certificates	true
scheduler.patch.image.registry	Certgen image registry	"docker.io"
scheduler.patch.image.repository	Certgen image repository	"jettech/kube-webhook-certgen"
scheduler.patch.image.tag	Certgen image tag	"v1.5.2"
scheduler.patch.image.pullPolicy	Certgen image pull policy	IfNotPresent
scheduler.patch.imageNew.registry	New certgen image registry	"docker.io"
scheduler.patch.imageNew.repository	New certgen image repository	"liangjw/kube-webhook-certgen"
scheduler.patch.imageNew.tag	New certgen image tag	"v1.1.1"

Scheduler Service Configuration

Parameter	Description	Default Value
scheduler.service.type	Service type	NodePort
scheduler.service.httpPort	HTTP port	443
scheduler.service.schedulerPort	Scheduler NodePort	31998
scheduler.service.monitorPort	Monitor port	31993
scheduler.service.monitorTargetPort	Monitor target port	9395

Device Plugin Configuration

Parameter	Description	Default Value
devicePlugin.image.registry	Device plugin image registry	"docker.io"
devicePlugin.image.repository	Device plugin image repository	"projecthami/hami"
devicePlugin.image.tag	Device plugin image tag	""
devicePlugin.image.pullPolicy	Device plugin image pull policy	IfNotPresent
devicePlugin.image.pullSecrets	Device plugin image pull secrets	[]

Monitor Configuration

Parameter	Description	Default Value
devicePlugin.monitor.image.registry	Monitor image registry	"docker.io"
devicePlugin.monitor.image.repository	Monitor image repository	"projecthami/hami"
devicePlugin.monitor.image.tag	Monitor image tag	""
devicePlugin.monitor.image.pullPolicy	Monitor image pull policy	IfNotPresent
devicePlugin.monitor.image.pullSecrets	Monitor image pull secrets	[]
devicePlugin.monitor.ctrPath	Container path	/usr/local/vgpu/containers
devicePlugin.monitor.extraArgs	Monitor extra arguments	["-v=4"]
devicePlugin.monitor.extraEnvs	Monitor extra environments	{}

Device Plugin Other Configuration

Parameter	Description	Default Value
devicePlugin.deviceSplitCount	Integer type, default value: 10. Maximum number of tasks assigned to a single GPU device	10
devicePlugin.deviceMemoryScaling	Device memory scaling ratio	1
devicePlugin.deviceCoreScaling	Device core scaling ratio	1
devicePlugin.runtimeClassName	Runtime class name	""
devicePlugin.createRuntimeClass	Whether to create runtime class	false
devicePlugin.migStrategy	String type, "none" means ignore MIG functionality, "mixed" means allocate MIG devices through independent resources	"none"
devicePlugin.disablecorelimit	String type, "true" means disable core limit, "false" means enable core limit	"false"
devicePlugin.passDeviceSpecsEnabled	Whether to enable passing device specs	false
devicePlugin.extraArgs	Device plugin extra arguments	["-v=4"]
devicePlugin.nodeConfiguration.config	Node configuration for device plugin by json	An example of default configuration.
devicePlugin.nodeConfiguration.externalConfigName	Node configuration for device plugin by external congimap	""
devicePlugin.extraEnvs	Device plugin extra environments	{}

Device Plugin Service Configuration

Parameter	Description	Default Value
devicePlugin.service.type	Service type	NodePort
devicePlugin.service.httpPort	HTTP port	31992

Device Plugin Deployment Configuration

Parameter	Description	Default Value
devicePlugin.pluginPath	Plugin path	/var/lib/kubelet/device-plugins
devicePlugin.libPath	Library path	/usr/local/vgpu
devicePlugin.nvidiaNodeSelector	NVIDIA node selector	{"gpu": "on"}
devicePlugin.updateStrategy.type	Update strategy type	RollingUpdate
devicePlugin.updateStrategy.rollingUpdate.maxUnavailable	Maximum unavailable count	1

Device Configuration

AWS Neuron

Parameter	Description	Default Value
devices.awsneuron.customresources	Custom resources	["aws.amazon.com/neuron", "aws.amazon.com/neuroncore"]

Kunlunxin

Parameter	Description	Default Value
devices.kunlun.enabled	Whether to enable	true
devices.kunlun.customresources	Custom resources	["kunlunxin.com/xpu"]

Mthreads

Parameter	Description	Default Value
devices.mthreads.enabled	Whether to enable	true
devices.mthreads.customresources	Custom resources	["mthreads.com/vgpu"]

NVIDIA

Parameter	Description	Default Value
devices.nvidia.gpuCorePolicy	GPU core policy	default
devices.nvidia.libCudaLogLevel	CUDA library log level	1

Huawei Ascend

Parameter	Description	Default Value
devices.ascend.enabled	Whether to enable	false
devices.ascend.image	Image	""
devices.ascend.imagePullPolicy	Image pull policy	IfNotPresent
devices.ascend.extraArgs	Extra arguments	[]
devices.ascend.nodeSelector	Node selector	{"ascend": "on"}
devices.ascend.tolerations	Tolerations	[]
devices.ascend.customresources	Custom resources	["huawei.com/Ascend910A", "huawei.com/Ascend910A-memory", ...]

Iluvatar

Parameter	Description	Default Value
devices.iluvatar.enabled	Whether to enable	false
devices.iluvatar.customresources	Custom resources	["iluvatar.ai/BI-V150-vgpu", "iluvatar.ai/BI-V150.vMem","iluvatar.ai/BI-V150.vCore", ...]