retry in case of empty response while fetching resource

[harshavmb]

This PR is an additional retry function with azapi_resource data provider.

The use case for this retry logic is that we have few Azure policies to update DNS configuration of private endpoints created by various users. There is some slight delay between the private endpoint resource creation & the Azure policy to create A-record & update the private endpoint DNS configuration. We have seen the time lag of up to 30-40 seconds.

Currently, we are using either null_resource & sleep providers. We wait deliberately for certain time expecting the configuration to be updated. Nevertheless, it's not a clean solution. We create inherent dependency on az-cli.

With this retry logic, azapi_resource data provider retries until non empty response is returned by the Azure API.

I know retrying bluntly on empty response isn't wise but I'm open to your suggestions.

[harshavmb]

@microsoft-github-policy-service agree

@microsoft-github-policy-service agree

[harshavmb]

Hi,

Any update on this PR?

[ms-henglu]

Hi @harshavmb ,

Thank you so much for taking time to work on this feature! And apologize for the late response.

We'll think about this feature request and improve it.

[ms-henglu]

We received a related feature request: #372

[harshavmb]

Hi @ms-henglu ,

I'm afraid the feature request you linked isn't the one this PR is about. It's not the resource, it's a data provider.
As I explained in the PR description, the create part is handled by Azure policies which have few seconds of latency & with data providers we would like to retry until the ARM resource is created by the policy.

[matt-FFFFFF]

I think #372 could cover both data sources and resources, using a common retry function.

[matt-FFFFFF]

Although, it might be a different scenario if the API does not return an error (which, I think it should)

[harshavmb]

Hi @matt-FFFFFF ,

Indeed if you prefer to have a common function, both resources & data sources could use that but I would like to have this available with data sources too as we have different use-cases than the one mentioned in #372.

[MatLambethMS]

Hi folks. How is this one looking? This feature would be extremely useful and looks like a lot of the hard work has been done.