Github user viirya commented on a diff in the pull request: https://github.com/apache/spark/pull/16700#discussion_r98325213 --- Diff: sql/hive/src/main/scala/org/apache/spark/sql/hive/HiveExternalCatalog.scala --- @@ -899,6 +919,21 @@ private[spark] class HiveExternalCatalog(conf: SparkConf, hadoopConf: Configurat spec, partitionColumnNames, tablePath) try { tablePath.getFileSystem(hadoopConf).rename(wrongPath, rightPath) + + // If the newSpec contains more than one depth partition, FileSystem.rename just deletes + // the leaf(i.e. wrongPath), we should check if wrongPath's parents need to be deleted. + // For example, give a newSpec 'A=1/B=2', after calling Hive's client.renamePartitions, + // the location path in FileSystem is changed to 'a=1/b=2', which is wrongPath, then + // although we renamed it to 'A=1/B=2', 'a=1/b=2' in FileSystem is deleted, but 'a=1' + // is still exists, which we also need to delete + val delHivePartPathAfterRename = getExtraPartPathCreatedByHive( --- End diff -- `client.renamePartitions` is called at the beginning of `renamePartitions` for all specs at once. It creates the directory `a=1` and `a=1/b=2` and `a=1/b=3`. When you iterates specs and rename the directories with FileSystem.rename, in the first iteration, `a=1/b=2` is renamed, and `a=1` is deleted in this change, then `a=1/b=3` will be deleted too. So in next iteration, the renaming of `a=1/b=3` to `A=1/B=3` will fail.
--- If your project is set up for it, you can reply to this email and have your reply appear on GitHub as well. If your project does not have this feature enabled and wishes so, or if the feature is enabled but not working, please contact infrastructure at infrastruct...@apache.org or file a JIRA ticket with INFRA. --- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org