Polling fails due "Request limit exceeded" and orphans instances

[sinneduy]

It looks like (PR) added retry functionality to handle "Request limit exceeded" for instances starting, and verifying volumes are up, but it doesn't have that for polling for an instance to exist.

Waited 0/300s for spot request <redacted> to become fulfilled.
--
  | Waited 5/300s for spot request <redacted> to become fulfilled.
  | Instance <redacted> requested.
  | Polling AWS for existence, attempt 0...
  | -----> Destroying <redacted>...
  | Finished destroying <redacted> (0m0.00s).

rake aborted!
--
  | Kitchen::ActionFailed: Failed to complete #create action: [stopped waiting due to an unexpected error: Request limit exceeded.]

Perhaps the wait_with_destroy call needs to be used for this?

Also, it looks like the instance doesn't actually get destroyed after this failure.

[ghost]

Have you tried "retry_limit"? https://github.com/test-kitchen/kitchen-ec2/blob/65796302b738152eb775ab88bae0531b42986b88/lib/kitchen/driver/aws/client.rb. This is not documented but appears to be passed directly to the AWS SDK.

[sinneduy]

that can reduce the frequency of this error, but it can still result in orphaned instances, if there are any failures...?

[xblitz]

Im also running into this issue... any plans to fix?