What happened:

After installing HAMI 2.7.0 using Helm, the hami-scheduler pod encounters RBAC permission errors preventing normal leader election and scheduling operations. Specific errors include:

1. User "system:serviceaccount:kube-system:hami-scheduler" cannot get resource "endpoints" in API group "" in the namespace "kube-system"
2. User "system:serviceaccount:kube-system:hami-scheduler" cannot list resource "configmaps" in API group "" in the namespace "kube-system"
3. Post "191": dial tcp 127.0.0.1:443: connect: connection refused

What you expected to happen:

HAMI scheduler should start normally, successfully perform leader election, and schedule GPU workloads to cluster nodes without permission errors.

**How to reproduce it **:

1. Install HAMI 2.7.0 using Helm:

   helm install hami . --set resourceName=[nvidia.com/vgpu](http://nvidia.com/vgpu) --set scheduler.kubeScheduler.imageTag=v1.19.6 -n kube-system

2. Check hami-scheduler pod logs:

   kubectl logs -f -l app=hami-scheduler -n kube-system

3. Observe the RBAC permission errors preventing normal operation

4. Insufficient RBAC permissions prevent access to required Kubernetes resources

Environment:

• HAMi version: 2.7.0
• nvidia driver or other AI device driver version: [Please provide nvidia-smi -a output]
• Docker version from docker version: [Please provide]
• Docker command, image and tag used: Installed via Helm chart with default images
• Kernel version from uname -a: [Please provide]
• Kubernetes version: v1.19.6