chihsuan commented on PR #10651: URL: https://github.com/apache/ozone/pull/10651#issuecomment-5267010605
Thanks again for the review! @yandrey321 @chungen0126 I’ve updated the PR with the following changes: - replaced MD5 with streaming CRC32 validation - aligned `--size`, `--buffer`, and `--copy-buffer` defaults to 16 KB - added the thread ID to each write marker so content is distinct across threads and overwrites - bounded per-thread history with `--max-files-per-thread` and path reuse - moved option validation before initialization and handled failed writes correctly - added integration coverage for FSO, LEGACY, path wrapping, and overwritten paths The remaining question appears to be the intended scope of this workload: 1. Is the validation requirement in HDDS-14524 still desired for Freon? If not, I think we should close this PR. 2. If validation is desired, should we use deterministic content with no per-path history, or keep the current approach? My preference is the current approach. Deterministic content without per-path state cannot detect stale reads when an overwritten path contains the same bytes. The current implementation uses a distinct marker for each write and retains the latest CRC32, with memory bounded by `--max-files-per-thread`. I’m happy to adjust based on consensus. Please let me know if I’m overlooking a simpler way to preserve stale-read detection. 🙏 -- 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]
