Since there are checks for submitting an invalid supervisor spec (e.g., if nothing in "bootstrap.servers" is resolvable), the error condition this is addressing seems to be situations where some external state changes and the supervisor spec no longer works. For the unresolvable bootstrap.servers issue as an example, maybe its a transient DNS error, so I feel like it's better to retry indefinitely without increasing backoff times (with retries tied to the configured supervisor run period instead). Otherwise, suppose the backoff time became extremely large but the user had fixed the underlying issue in their environment, with a huge backoff time they would have to restart the overlord for the change to take effect.
RetryUtils.retry() is also a blocking call, which I want to avoid here in the lifecycle startup. [ Full content available at: https://github.com/apache/incubator-druid/pull/6383 ] This message was relayed via gitbox.apache.org for [email protected]
