len548 opened a new pull request, #12065: URL: https://github.com/apache/ozone/pull/12065
## What changes were proposed in this pull request? PutBlock reply handlers of one Ratis `BlockOutputStream` can run out of order. The Ratis client completes the replies in order, but thenApplyAsync runs each handler on the shared responseExecutor, so the handler of a later PutBlock can run before the handler of an earlier one. That handler moves the bcsId backward, and close() can send OM the full block length with the older bcsId. If a later request fails first, `CommitWatcher` can also acknowledge the later buffers while the earlier data is still allocated, and the retry rewrites that earlier data. Each PutBlock handler, including a piggybacked PutBlock on WriteChunk, now waits with thenCombineAsync until the previous handler of that stream has finished. The handlers run in send order, so the bcsId only moves forward and the acknowledged buffers stay a prefix of the block. ## What is the link to the Apache JIRA https://issues.apache.org/jira/browse/HDDS-17668 ## How was this patch tested? Added two tests in `TestBlockOutputStreamCorrectness`. The tests queue the handlers and run the newest one first. One checks that the bcsId stays at the last PutBlock. The other fails a later write and checks that the buffers still allocated are the bytes after the acknowledged length. -- 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]
