Resource Pool Manager Chart 0.0.6
Chart: 0.0.6 · appVersion: 0.0.2 · 发布时间: 2026-09-23 (UTC)
软件变更
RPM 0.0.2 新增 Hygon HCU 设备清单、资源池选择和分配准入支持。manager 从已有 HAMi scheduler 设备配置中读取 Hygon 资源名称。
NVIDIA、Ascend 和 Hygon 后端可分别启用,至少需要启用一类后端。设备配置须与 Webhook 匹配的资源一致。
Chart 变更
appVersion从0.0.1升级为0.0.2,默认镜像改为ghcr.io/dynamia-ai/resource-pool-manager:v0.0.2。devices.nvidia.enabled默认为true,devices.ascend.enabled和devices.hygon.enabled默认为false。hami.ascendEnabled替换为devices.ascend.enabled。- Webhook TLS 沿用 Chart 0.0.5 引入的
webhook.tls配置。
升级说明
从现有 values 中删除 hami.ascendEnabled,将原值迁移到 devices.ascend.enabled。使用 Hygon HCU 时,启用 devices.hygon.enabled,并确认 scheduler 设备配置包含对应的 HCU 资源名称。只使用 Ascend 或 Hygon 的安装需显式关闭 NVIDIA。
同时升级 Chart 与 manager 镜像。自定义 image.tag 不随 appVersion 变化,须显式更新。设备资源名或已启用后端变化时,Webhook 规则与服务副本需要完成切换,期间可能暂时拒绝新建 Pod 的准入请求。
仍要求 Kubernetes 1.30 或更新版本、单独安装的资源池 CRD、HAMi extender 调度模式,以及指向已有 scheduler 设备 ConfigMap 的 hami.existingConfigMap。修改 scheduler 设备配置后,使用完整 manager values 文件升级 RPM,重新生成 Webhook 规则与 manager 参数。
Values 说明
以下参数说明取自 Chart 0.0.6 发布包随附的 README,保留原文。默认配置以该版本的 values.yaml 为准。
展开参数表
| Key | Type | Default | Description |
|---|---|---|---|
| config.create | bool | true | Initialize the runtime ConfigMap if absent. It is retained outside Helm ownership for external policy updates. For render-and-apply GitOps, precreate the ConfigMap and set this to false. |
| config.name | string | "" | Runtime ConfigMap name; defaults to <workload name>-config. With config.create=false it must already exist in the manager namespace. |
| config.poolSchedulerPolicy | string | "binpack" | Initial pool selection policy. Later changes belong in the runtime ConfigMap, not Helm values. |
| devices.ascend.enabled | bool | false | Include Ascend profiles from the scheduler device ConfigMap in resource-pool admission. |
| devices.hygon.enabled | bool | false | Include Hygon HCU resources from the scheduler device ConfigMap in resource-pool admission. |
| devices.nvidia.enabled | bool | true | Include NVIDIA resources from the scheduler device ConfigMap in resource-pool admission. |
| fullnameOverride | string | "" | Override the workload name. Names are limited to 43 characters to leave room for resource suffixes. |
| hami.existingConfigMap | string | "" | Required name of the scheduler device ConfigMap in this release namespace. Helm reads data.device-config.yaml from the cluster and the manager mounts the same ConfigMap. Inline configuration and overrides are not supported. Upgrade the manager after scheduler device configuration changes. |
| hami.schedulerName | string | "hami-scheduler" | Scheduler name used by the existing HAMi installation. The scheduler must use HAMi extender admission and scheduling, not DRA. |
| healthProbeBindAddress | string | ":8081" | Health server listen address. |
| image.pullPolicy | string | "IfNotPresent" | Resource-pool-manager image pull policy. |
| image.repository | string | "ghcr.io/dynamia-ai/resource-pool-manager" | Resource-pool-manager image repository, including registry. |
| image.tag | string | "v0.0.2" | Required published image tag. Independent of this chart's appVersion; update explicitly when releasing a new manager image. |
| imagePullSecrets | list | [] | Registry Secret names used by the manager and certificate Jobs. |
| leaderElect | bool | true | Elect a leader for inventory and status reconcilers. |
| metricsBindAddress | string | ":8080" | Metrics server listen address. |
| nameOverride | string | "" | Override the chart name used in names and labels. |
| nodeSelector | object | {} | Node selector for the manager. |
| podAnnotations | object | {} | Additional annotations for manager Pods. |
| replicas | int | 1 | Number of manager replicas. Use leader election when running multiple replicas. |
| resources | object | {} | Resource requests and limits for the manager. |
| timeoutSeconds | int | 10 | Admission webhook timeout in seconds. |
| tolerations | list | [] | Tolerations for the manager. |
| webhook.tls.certManager | bool | false | Use an installed cert-manager controller for certificates and CA injection. When false, certgen hook Jobs manage webhook certificates. |
| webhook.tls.certgen.image.pullPolicy | string | "IfNotPresent" | Certificate generator image pull policy. |
| webhook.tls.certgen.image.repository | string | "registry.k8s.io/ingress-nginx/kube-webhook-certgen" | Certificate generator image repository, including registry. |
| webhook.tls.certgen.image.tag | string | "v1.6.9" | Certificate generator image tag. |
| webhook.tls.certgen.nodeSelector | object | {} | Node selector for certificate Jobs. |
| webhook.tls.certgen.podAnnotations | object | {} | Additional annotations for certificate Job Pods. |
| webhook.tls.certgen.priorityClassName | string | "" | Priority class for certificate Jobs. |
| webhook.tls.certgen.runAsUser | int | 65532 | Non-root UID for certificate Jobs. |
| webhook.tls.certgen.tolerations | list | [] | Tolerations for certificate Jobs. |
| webhookPort | int | 9443 | HTTPS port in the manager container. |