Machine with error`ZONE_RESOURCE_POOL_EXHAUSTED_WITH_DETAILS` is not flagged as user error

[AleksandarSavchev]

How to categorize this issue?

/area ops-productivity /kind bug /platform gcp

What happened: With @ialidzhikov we found that an machine error such as

status:
  currentStatus:
    lastUpdateTime: "2024-05-21T12:57:28Z"
    phase: CrashLoopBackOff
  lastOperation:
    description: 'Cloud provider message - machine codes error: code = [ResourceExhausted]
      message = [Create machine "machine"
      failed: The zone ''zone'' does not have
      enough resources available to fulfill the request.  ''(resource type:compute)''.]'
    errorCode: ResourceExhausted
    lastUpdateTime: "2024-05-21T12:57:28Z"
    state: Failed
    type: Create

is not properly categorised as user error since it should be matched by https://github.com/gardener/gardener-extension-provider-gcp/blob/c32ba9eec28e676d15de9c55b08b78ff0235c0b5/pkg/apis/gcp/helper/error_codes.go#L16

however ZONE_RESOURCE_POOL_EXHAUSTED_WITH_DETAILS is replaced by ResourceExhausted here: https://github.com/gardener/machine-controller-manager-provider-gcp/blob/295ac09467c51746f87762130a25934be202df68/pkg/gcp/machine_controller_util.go#L440-L446

What you expected to happen: Error to be properly flagged as user error.

How to reproduce it (as minimally and precisely as possible):

Anything else we need to know?:

Environment:

• Gardener version (if relevant):
• Extension version: v1.35.0
• Kubernetes version (use kubectl version):
• Cloud provider or hardware configuration:
• Others:

[kon-angelo]

Do you think it makes sense if instead of fixing this occurrence, we rather make the extension worker library aware of MCM error codes and map them to gardener error codes ? WDYT @ialidzhikov ?

[ialidzhikov]

Sounds reasonable, at least for the machine-controller-manager error codes that we can map unambiguously - Unauthenticated, PermissionDenied, ResourceExhausted. We can even do it without regex, as the corresponding error code is present in the Machine status (.status.errorCode field). I didn't check whether it is propagated to the MachineDeployment status.