javier commented on PR #4505: URL: https://github.com/apache/iggy/pull/4505#issuecomment-6145173176
Thanks for the review. Pushed fixes for everything except the consume-error point, which I replied to inline. Code: - The chunk loop drains a VecDeque now, so a batch bigger than `batch_size` no longer shifts the queue on every chunk. - The rejected-payload preview writes JSON through a capped writer instead of rendering the whole document first. - Bad config values now say what was wrong and what's accepted, and zero `batch_size` / `max_flush_bytes` log a warning instead of silently falling back. - The row path sizes its vectors up front. README: - Delivery is still at least once, but the section now lists the three ways rows can be lost: `ack_level = "ok"` on a cluster, no `sf_dir` on a crash or slow shutdown, and records the sink rejects. - Documented the oversized single row case, that `timestamp_unit` only applies to payload timestamps, and how a late terminal rejection can fail another topic's batch. - While going through it I checked every connect-string key against questdb-rs and fixed two mistakes: the key is `close_flush_timeout_millis`, not `close_flush_timeout`, and the frame cap comes from `sf_max_segment_bytes` with the defaults, not `max_buf_size`. Clippy, unit tests, prek and the testcontainers suite all pass. -- 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]
