Description of problem:

In openshift/installer, when running GetMarketplaceImage, we get 

encountered error: failed to fetch marketplace image: could not get marketplace image: compute.VirtualMachineImagesClient#Get: Failure sending request: StatusCode=0 – Original Error: context deadline exceeded

This error comes from https://github.com/openshift/installer/blob/2b773d8339292cb3f2ba294e3b8f186da9cf7395/pkg/asset/installconfig/azure/client.go#L367-L378

However, it doesn't have to timeout in 30 secs here, and considering the time required to the whole install, we should retry the request when it fails decent number of times.

Also it's still using the deprecated SDK, which lacks retry mechanism.

With the new SDK, there's a rich retry mechanism and that should retry when it gets certain errors.

https://learn.microsoft.com/en-us/azure/storage/blobs/storage-retry-policy-go

https://azure.github.io/azure-sdk/releases/deprecated/go.html

https://github.com/Azure/azure-sdk-for-go/blob/69f43460d64e153c91e6fe52113ad31ce6f91a06/sdk/azcore/runtime/policy_retry.go#L106

Version-Release number of selected component (if applicable):

all versions implementing GetMarketplaceImage

How reproducible:

when there's a network issue.

Steps to Reproduce:

It relates to the network issue. not reproducible always.

    1.
    2.
    3.
    

Actual results:

It terminates immediately.

Expected results:

It should retry some times.

Additional info: