diggle1 commented on PR #99: URL: https://github.com/apache/doris-kafka-connector/pull/99#issuecomment-5099401410
Thanks for the review — you're right. I realize the previous sticky TTL cache changed the per-call round-robin selection, and in some cases it could skew load onto a single BE within the cache window. That was not what I intended. I've updated the patch as follows: 1. I keep round-robin selection with `pos++` on every `getAvailableBackend()` call, so load remains balanced across BEs. 2. I only cache per-BE HTTP probe success for a short TTL (5s). If a BE was recently confirmed alive, I skip `tryHttpConnection()` for that BE on the next selection. 3. On stream-load / commit failure, or when no BE is available, I clear the probe cache via `invalidateCache()`, so the next attempt re-probes instead of trusting a stale result. Please take another look when you have time. -- 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]
