发生了什么？

当使用 ResourceQuotas 准入控制器扩展资源( nvidia.com/gpu )时，报告的已使用资源不一致且错误。

你期望会发生什么？

报告的已使用扩展资源应尊重命名空间中的当前状态。

我们如何尽可能最小精确地重现它？

应用以下 ResourceQuota:

apiVersion: v1
kind: ResourceQuota
metadata:
  name: capsule-skg00000-2
spec:
  hard:
    requests.nvidia.com/gpu: "10"

检查 RQ 是否正确应用并报告预期状态：

apiVersion: v1
kind: ResourceQuota
metadata:
  annotations:
    kubectl.kubernetes.io/last-applied-configuration: |
{"apiVersion":"v1","kind":"ResourceQuota","metadata":{"annotations":{},"name":"capsule-skg00000-2","namespace":"skg00000-gpu"},"spec":{"hard":{"requests.nvidia.com/gpu":"10"}}}
  creationTimestamp: "2024-05-26T17:00:28Z"
  labels:
    capsule.clastix.io/managed-by: skg00000
  name: capsule-skg00000-2
  namespace: skg00000-gpu
  resourceVersion: "8225530"
  uid: 3badf2c2-9959-4882-9315-7ffff127e08e
spec:
  hard:
    requests.nvidia.com/gpu: "10"
status:
  hard:
    requests.nvidia.com/gpu: "10"
  used:
    requests.nvidia.com/gpu: "0"

创建一个使用扩展资源(GPU)的模板的部署。

spec:        
  automountServiceAccountToken: false   
  containers:
  - args:    
    - sleep 3600      
    command: 
    - /bin/bash       
    - -c     
    - --     
    image: nvidia/cuda:11.0.3-base-ubuntu20.04   
    imagePullPolicy: Always    
    name: nvidia      
    resources:        
      limits:
        cpu: "1"      
        memory: 1Gi   
        nvidia.com/gpu: "1"    
      requests:       
        cpu: "1"      
        memory: 1Gi   
        nvidia.com/gpu: "1"

在 Pod 部署后，ResourceQuota 报告错误的已使用资源数量：

apiVersion: v1
kind: ResourceQuota
metadata:
  annotations:
    kubectl.kubernetes.io/last-applied-configuration: |
{"apiVersion":"v1","kind":"ResourceQuota","metadata":{"annotations":{},"name":"capsule-skg00000-2","namespace":"skg00000-gpu"},"spec":{"hard":{"requests.nvidia.com/gpu":"10"}}}
  creationTimestamp: "2024-05-26T17:00:28Z"
  labels:
    capsule.clastix.io/managed-by: skg00000
  name: capsule-skg00000-2
  namespace: skg00000-gpu
  resourceVersion: "8225807"
  uid: 3badf2c2-9959-4882-9315-7ffff127e08e
spec:
  hard:
    requests.nvidia.com/gpu: "10"
status:
  hard:
    requests.nvidia.com/gpu: "10"
  used:
    requests.nvidia.com/gpu: "2"

执行 Deployment 的扩展时也会出现不一致的报告。

$: kubectl scale deployment gpu-workload --replicas=2
$ RQ reports 4 used instances
$: kubectl scale deployment gpu-workload --replicas=3
$ RQ reports 6 used instances
$: kubectl scale deployment gpu-workload --replicas=4
$ RQ reports 8 used instances

在缩减规模时，使用的示例仍会不时出现不一致的情况。

$: kubectl scale deployment gpu-workload --replicas=3
$ RQ reports 3 used instances
$: kubectl scale deployment gpu-workload --replicas=2
$ RQ reports 3 used instances
$: kubectl scale deployment gpu-workload --replicas=1
$ RQ reports 1 used instances
$: kubectl scale deployment gpu-workload --replicas=0
$ RQ reports 0 used instances

我们需要了解其他信息吗？

• 无响应*

Kubernetes 版本

$ kubectl version
Client Version: v1.30.0
Kustomize Version: v5.0.4-0.20230601165947-6ce0bf390ce3
Server Version: v1.29.4

云提供商

OS 版本

# On Linux:
$ cat /etc/os-release
PRETTY_NAME="Ubuntu 24.04 LTS"
NAME="Ubuntu"
VERSION_ID="24.04"
VERSION="24.04 LTS (Noble Numbat)"
VERSION_CODENAME=noble
ID=ubuntu
ID_LIKE=debian
HOME_URL="https://www.ubuntu.com/"
SUPPORT_URL="https://help.ubuntu.com/"
BUG_REPORT_URL="https://bugs.launchpad.net/ubuntu/"
PRIVACY_POLICY_URL="https://www.ubuntu.com/legal/terms-and-policies/privacy-policy"
UBUNTU_CODENAME=noble
LOGO=ubuntu-logo
$ uname -a
Linux REDACTED 6.8.0-31-generic #31-Ubuntu SMP PREEMPT_DYNAMIC Sat Apr 20 00:40:06 UTC 2024 x86_64 x86_64 x86_64 GNU/Linux

安装工具

kubeadm

容器运行时(CRI)和版本(如适用)

N.R.

相关插件(CNI,CSI,...)和版本(如适用)

N.R.