Http 400 Preflight validation check error the operation failed or was cancelled needs detailed message

[hemarina]

Bug Report

• import path of package in question: preflight validate
• SDK version: armresources, latest, 1.2.0
  • Specify the exact commit if possible: b3194dd#diff-201f295ddf411fc3a73ca555f5a3da5f3796d1dff4075ec894d07292bc79fc1b
• output of go version: go1.22.1 windows/amd64

I'm testing preflight validation function with an unsupported location (e.g australiacentral) for todo-nodejs-mongo-aks. The error message is the operation failed or was cancelled. The error message should provide a detailed message on why it is failing.

I debugged and this is the error message that got skipped: The template deployment '****' is not valid according to the validation procedure. The tracking id is '****'. See inner errors for details. Preflight validation check for resource(s) for container service *** in resource group ****** failed. Message: Virtual Machine size: '' is not supported for subscription ***** in location 'australiacentral'. Please refer to aka.ms/aks/vm-size-selector to find supported VM sizes in location 'australiacentral'. This message is necessary for users to debug and should be provided in error message.

[tadelesh]

could you provide your go code of using the BeginValdate method?

[github-actions]

Hi @hemarina. Thank you for opening this issue and giving us the opportunity to assist. To help our team better understand your issue and the details of your scenario please provide a response to the question asked above or the information requested above. This will help us more accurately address your issue.

[hemarina]

@tadelesh

func (ds *deployments) ValidatePreflightToResourceGroup(
	ctx context.Context,
	subscriptionId, resourceGroup, deploymentName string,
	armTemplate azure.RawArmTemplate,
	parameters azure.ArmParameters,
	tags map[string]*string,
) (*armresources.DeploymentPropertiesExtended, error) {
	deploymentClient, err := ds.createDeploymentsClient(ctx, subscriptionId)
	if err != nil {
		return nil, fmt.Errorf("creating deployments client: %w", err)
	}

	validate, err := deploymentClient.BeginValidate(ctx, resourceGroup, deploymentName,
		armresources.Deployment{
			Properties: &armresources.DeploymentProperties{
				Template:   armTemplate,
				Parameters: parameters,
				Mode:       to.Ptr(armresources.DeploymentModeIncremental),
			},
			Tags: tags,
		}, nil)
	if err != nil {
		return nil, fmt.Errorf("calling preflight validate api failing: %w", err)
	}

	validateResult, err := validate.PollUntilDone(ctx, nil)
	if err != nil {
		deploymentError := createDeploymentError(err)
		return nil, fmt.Errorf(
			"validating preflight to resource group:\n\nDeployment Error Details:\n%w",
			deploymentError,
		)
	}

	return validateResult.DeploymentValidateResult.Properties, nil
}

https://github.com/hemarina/azure-dev/blob/4b4f6a3779d40b9d95914b26a28d63688da96edb/cli/azd/pkg/azapi/deployments.go#L296

I got error message: ERROR: deployment failed: error deploying infrastructure: calling preflight validate api failing: the operation failed or was cancelled

[tadelesh]

refer the following code to deal with error code and error message

var rawResponse *http.Response
ctx := context.TODO() // your context
ctxWithResp := runtime.WithCaptureResponse(ctx, &rawResponse)
resp, err := resourceGroupsClient.CreateOrUpdate(ctxWithResp, resourceGroupName, resourceGroupParameters, nil)
if err != nil {
    // with error, you can get RawResponse from context
    log.Printf("Status code: %d", rawResponse.StatusCode)
    var respErr *azcore.ResponseError
    if errors.As(err, &respErr) {
        // with error, you can also get RawResponse from error
        log.Fatalf("Status code: %d", respErr.RawResponse.StatusCode)
    } else {
        log.Fatalf("Other error: %+v", err)
    }
}
// without error, you can get RawResponse from context
log.Printf("Status code: %d", rawResponse.StatusCode)

[weikanglim]

I can add some debugging notes here since I was fortunate enough to be present with @hemarina at the time, but I do believe this is a bug in the Swagger spec that is used to generate deployment_clients.go.

Here's the execution flow we observed:

1. Client calls BeginValidate in deployment_clients.go.
2. ARM returns 400 Bad Request
3. validate checks for Bad Request, but considers it a success state, it returns httpResp, nil instead of nil, err.
4. Execution continues until NewPoller, where it's rejected by the validation check.

In summary, the bug observed by the user is:

• When ARM returns 400 bad request for deployments/validate, the error the operation failed or was cancelled is returned instead of the expected azcore.ResponseError.

We are also happy to provide a minimal repro setup, but it boils down to calling deploymentsClient.BeginValidate with parameters that will result in an ARM 400 response.

[tadelesh]

@weikanglim thanks for the detail. if possible, could you provide the detail logs with this instruction. i want to confirm the service return 400 with the lro init call.

[rajeshkamal5050]

@hemarina can you enable logging and share?

[hemarina]

POST https://management.azure.com/subscriptions/***/resourcegroups/***providers/Microsoft.Resources/deployments/***/validate?api-version=2021-04-01
   Accept: application/json
   Authorization: REDACTED
   Content-Length: 10435
   Content-Type: application/json
   User-Agent: azsdk-go-armresources.DeploymentsClient/v1.1.1 (go1.22.1; Windows_NT),azdev/0.0.0-dev.0 (Go go1.22.1; windows/amd64)
   X-Ms-Correlation-Request-Id: ***
   --------------------------------------------------------------------------------
   RESPONSE Status: 400 Bad Request
... Detailed version sent through Teams...

Retry: response 400
Retry: exit due to non-retriable status code

Status code: 400
Other error: the operation failed or was cancelled

[tadelesh]

@jhendrixMSFT For LRO with 400 response definition, could it be the initial request response result? (see spec here)
@hemarina Before we confirm if it is a service issue, I think you could also try to get the error info with my comment code with the first call of LRO, it should also work for LRO.

[jhendrixMSFT]

Including a 400 in the responses indicates it should be treated as a success response. However, if you look at the schema for the 400 it's an error type. The service team either needs to remove the 400 or add x-ms-error-response to it.

[tadelesh]

Got it. Thanks Joel. Let me add service attention label.

[github-actions]

Thanks for the feedback! We are routing this to the appropriate team for follow-up. cc @armleads-azure.

[hemarina]

@tadelesh @armleads-azure @jhendrixMSFT azd <-> preflight integration is currently blocked due to this.

Could you prioritize fixing this issue and provide an ETA?

[tadelesh]

Our SDK is auto-generated from service's spec. @armleads-azure could you help to prioritize?