Hi Hadoop developers, I'd like to request a review of a fix for HADOOP-17611, where DistCp's parallel file copy (-blocksperchunk) breaks modification time preservation.
When -pt is specified, the mapper preserves timestamps on the chunks, but the subsequent concat operation changes the final file's modification time. This patch restores the source modification and access times after concatenation and checksum validation, only when TIMES preservation is requested. The PR adds 14 HDFS contract test cases covering timestamp preservation enabled/disabled, ordinary/direct/atomic copies, directory scope, and update/skip behavior, as well as checks for other preserved attributes. The targeted regression suite completed with 83 tests passed, 16 skipped, and no failures or errors. The original implementation fails the three TIMES-enabled copy-mode cases, which pass with the fix. JIRA: https://issues.apache.org/jira/browse/HADOOP-17611 PR: https://github.com/apache/hadoop/pull/8858 Feedback and reviews would be appreciated. Thanks, Xinyu Tan --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
