georgew5656 commented on PR #17509: URL: https://github.com/apache/druid/pull/17509#issuecomment-2529112013
> > i thought about this approach but the concern is that it's possible for these error logs (inconsistency with metadata state) to be final error states, like if offsets are manually updated or something. we don't really want to retry in these scenarios > > Hmm, I don't think I have ever seen a case of people updating offsets manually. At best, someone might completely remove the entry for a certain datasource, in which case the flow would still work. > > And even for that, simply resetting the supervisor is advised (and preferred by users) rather than updating the DB directly. i ran some more testing and i've discovered that the scenario described in this PR initially only triggers the "The new start metadata state[%s] is ahead of the last committed" message. (Later tasks can fail with the inconsistent metadata state message). Based on this I think its fine to just throw a retryable exception in the first case and let the mysql connector retry instead of passing it up to the task action level (it looks like it defaults to 10 retries). I think this" new start metadata is ahead" state is relatively safe to retry as well, i believe the terminal failures states normally throw "inconsistent metadata" instead -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
