Open luiof opened 6 months ago
This shouldn't be too difficult to implement. The logic in requestConfig.MakeRequest() can be extended to retry requests that fail with a 429 error code. I think adding parameters for max retries, exponential backoff, and jitter should all help with the problem described here.
Additionally those parameters should be exposed through the provider config so they can be fine-tuned.
We are trying to create resources on the LogDNA provided by IBM and when the set of resources to create is huge, we get always a rate limited response.
The problem can be skipped in the apply since we can use some delays between the resources; but in the destroy the problem is still there and the terraform state is not consistent due to the rate limited responses.
This should be handled by the provider.