xiangfu0 commented on PR #18742: URL: https://github.com/apache/pinot/pull/18742#issuecomment-5975815156
[@Jackie-Jiang](https://github.com/Jackie-Jiang), you are right that the previous change tightened the best-efforts contract. I narrowed the guard to the last observed serving replica: it allows 2→1 with a configured minimum of 2, safe partial progress, and moves for segments with no serving replicas. Serving sources are selected before assignment calculation, and unready/ERROR targets can load while those sources are retained. Strict replica groups retain a common serving source where one exists. This still differs from `bestEfforts=false`, which stops on ERROR/nonconvergence. `FAILED` remains when no safe next assignment can advance: `DONE` would mark an incomplete target finished and suppress retries. `minAvailableReplicas=0` remains the explicit opt-out. Could you confirm whether this failure boundary is acceptable? Updated in [9bded0fc1e](https://github.com/apache/pinot/commit/9bded0fc1eb4dfc3c332f2ae9313c25b2776c026), rebased onto current master. All 24 targeted tests, formatting/checkstyle/license checks, and independent review passed. The standard local reactor hits an unchanged `ZstandardDecompressor` missing-`NotNull` dependency; the passing run used a compiler-only JetBrains annotations classpath supplement. Fresh hosted CI is separate from these local results. Re-review requested. -- 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]
