feat: adapt for gpu cost

[bxy4543]

🤖 Generated by Copilot at d5f5b71

Summary

🎮🛡️📊

This pull request adds support for Nvidia GPU resources in the sealos controller. It defines GPU-related constants, types, and functions in the gpu package, and uses them to create, price, and monitor GPU resources in the resources.go and monitor_controller.go files. It also updates the RBAC and deployment configurations for the monitor-controller controller to allow it to access node information.

We're the monitor-controller crew, we watch the nodes and pods
We gather GPU data and we store it in our logs
We heave and haul on the MonitorReconciler line
And we sing this shanty chorus on the count of three and nine

Walkthrough

• Add support for collecting and storing Nvidia GPU resource usage data from pods and nodes (link,link,link,link,link,link,link,link,link,link,link,link,link,link)
• Define constants, types, and functions related to Nvidia GPU labels and information in nvidia.go (link)
• Import the gpu package and add a new constant ResourceGPU and a new function NewGpuResource to resources.go (link,link)
• Add the ResourceGPU constant to the PricesUnit map in resources.go to define the unit price for GPU resources (link)
• Add a new rule to role.yaml and deploy.yaml to allow the monitor role and the monitor-controller controller to get, list, and watch nodes (link,link)
• Import the gpu package and add a new field NvidiaGpu to the MonitorReconciler struct in monitor_controller.go (link,link)
• Add a kubebuilder annotation to monitor_controller.go to generate RBAC rules for accessing nodes (link)
• Modify the NewMonitorReconciler function in monitor_controller.go to call the GetNodeGpuModel function from the gpu package and assign the result to the NvidiaGpu field (link)
• Modify the podResourceUsage function in monitor_controller.go to handle the GPU resource usage from pods by checking the GPU limit, getting the GPU model, and adding the GPU request to the rs map (link)
• Modify the getResourceValue function in monitor_controller.go to convert the GPU resource usage to integer values based on the unit price (link)
• Modify the initResources function in monitor_controller.go to initialize the rs map with zero values for the ResourceGPU constant (link)
• Add error handling and logging for the case when a user-owned namespace does not have a resource quota for storage (link)
  • Add a new variable hasStorageQuota to indicate whether the namespace has a resource quota for storage in the podResourceUsage function in monitor_controller.go (link)
  • Add a condition to check if the namespace has the UserAnnotationOwnerKey annotation and log an error if the resource quota is empty in the podResourceUsage function in monitor_controller.go (link)
  • Add a block of code to list all the PVCs in the namespace and add their storage requests to the rs map if the namespace does not have a storage quota in the podResourceUsage function in monitor_controller.go (link)

[sealos-ci-robot]

🤖 Generated by lychee action

Summary

Status	Count
🔍 Total	914
✅ Successful	347
⏳ Timeouts	0
🔀 Redirected	0
👻 Excluded	566
❓ Unknown	0
🚫 Errors	0

Full action output

Full Github Actions output

[codecov]

Codecov Report

Patch and project coverage have no change.

Comparison is base (0adc21d) 67.92% compared to head (d286b41) 67.92%.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #3596   +/-   ##
=======================================
  Coverage   67.92%   67.92%           
=======================================
  Files           8        8           
  Lines         664      664           
=======================================
  Hits          451      451           
  Misses        171      171           
  Partials       42       42           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.