Hmm, I think backoff is more appropriate for transient errors. It starts with retrying in a very short interval at first, but the retry interval grows as the failure continues. This makes sense because it can recover quickly with short retry periods if the error is transient. If the error is permanent, the retry period would grow in the end and it won't use excessive resources. Regarding too large backoff time, `RetryUtils.retry()` has the `MAX_SLEEP_MILLIS` which I think it makes sense to make it configurable. Regarding blocking call, please check my below comment. What do you think?
[ Full content available at: https://github.com/apache/incubator-druid/pull/6383 ] This message was relayed via gitbox.apache.org for [email protected]
