定义50 的虚拟卡,却能分配大于2个pod到一张卡上

[weapons97]

如下定义8个pod, 每个pod 分配volcano.sh/vgpu-cores: "50", 但是进入检查pod gpu bus-id,看到6个pod都在同一个卡上

apiVersion: apps/v1
kind: Deployment
metadata:
  name: xxx
spec:
  replicas: 8
  selector:
    matchLabels:
      app: cnn
  template:
    metadata:
      labels:
        app: cnn
    spec:
      schedulerName: volcano
          resources:
            limits:
              # volcano.sh/vgpu-cores: "100"
              # volcano.sh/vgpu-memory: "20000"
              # volcano.sh/vgpu-number: "1"
              # nvidia.com/gpu: 1
              volcano.sh/vgpu-cores: "50"
              volcano.sh/vgpu-memory: "4544"
              volcano.sh/vgpu-number: "1"
...

[archlitchi]

check curl {volcano scheduler cluster ip}:8080/metrics

[weapons97]

check curl {volcano scheduler cluster ip}:8080/metrics

可以看到有300 cores 的，这合理吗

[archlitchi]

which version of volcano are you using?

[weapons97]

which version of volcano are you using?

1.9.0

[archlitchi]

fixed, related PR:volcano-sh/volcano#3774

waiting for volcano-community to integrate