If a request is throttled by a cloud provider during Discovery, Cloud Request Retry Configuration provides a customizable method to retry requests. Discovery and Service Mapping Patterns includes a retry configuration for AWS and Azure. You can customize the included configuration or create your own.

Discovery admins and Cloud admins can access the request retry configuration at All > Discovery > Cloud Request Retry Configuration. You can create one configuration for each provider.

When a request is throttled, the retry framework uses retry configuration defined for the provider to handle retries before returning the final response to the ApiCommand classes:
  • AwsApiCommand
  • AzureApiCommand

The retry configurations are synced with the MID Servers through the MID Server property, mid.cloud.discovery.retry.configuration.

There are the following retry strategies:
  • Exponential backoff
  • Response header backoff
  • Custom backoff

Exponential backoff

For the following example configuration:
SettingValue
Max retries 3
Response codes 429
Base delay in ms 1000
Max delay in ms 10000
Additional delay window in ms 1500
The Exponential backoff retry strategy works as follows:
  • 1st retry—the backoff multiplier is randomly selected between 0 and 1. The max delay value is 400 ms (400 * 1).
  • 2nd retry—the backoff multiplier is randomly selected between 0 and 3. The max delay value is 1200 ms (400 * 3).
  • 3rd retry—the backoff multiplier is randomly selected between 0 and 7. The max delay value is 2800 ms (400 * 7).

On subsequent retries, if the delay exceeds 10000 (the max delay), 10000 will be used as initial delay.

Once the initial delay is generated, jitter is added to the delay. The jitter window is defined by the Additional delay window in ms field. The system selects a random value between 0 and 1500 and adds it to the initial delay.

If the initial delay is 500, the final delay (with jitter) can be a value between 500 and 2000 ms.

Response header backoff

For the following example configuration:
SettingValue
Max retries 3
Response codes 429
Response header Retry-After
Response header delay unit Seconds
Additional delay window in ms 1500
The Response header backoff strategy works as follows:
  • Fetch the value of header Retry-After from the server response.
  • Convert the Retry-After to milliseconds by multiplying by 1000.

Once the initial delay is generated, jitter is added to the delay. The jitter window is defined by the Additional delay window in ms field. The system selects a random value between 0 and 1500 and adds it to the initial delay.

If the initial delay is 2000, the final delay (with jitter) can be a value between 2000 and 3500 ms.

Custom backoff

With a custom backoff retry strategy, you define the Max retries and Response codes and create your own Mid script include that defines how requests are retried using the getDelay() function. For more information, see Script includes.