I am using --warehouse-dir argument for loading data in HDFS before sqoop puts it into hive. I am running all my sqoop jobs through oozie.
Now, if the task fails for some reason, it is reattempted and the problem here is that the warehouse dir created by previous task is still there and the task re-attempt fails with error : output directory already exists.
I understand I could use direct argument to skip intermediate loading in HDFS step but I need to use drop import hive delims argument as well and that's not supported with Hive. Advice, please? It's important.
This is a normal behavior.
You should either create dynamic folder name (e.g. output_dir_timestamp) but you may end up having a lot of directories, or add an HDFS action to delete the HDFS directory, just before the sqoop action. I recomend the last approach.