----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/11083/#review20491 -----------------------------------------------------------
I think we should refactor CliDriver.run into the setup and the part that actually runs the commands. If we pull out from the part where the cli object is created down, we can isolate all of the multiple exits to that routine and make the ss.close handling more future proof. In terms of changing the session id to a uuid, I think that it is better to have a human readable string than a random identifier. Since the current session id will be unique upto a process, maybe we could add a static counter that keeps track of how many session ids this process has created and add that as a suffix. - Owen O'Malley On May 13, 2013, 8:50 a.m., Prasad Mujumdar wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/11083/ > ----------------------------------------------------------- > > (Updated May 13, 2013, 8:50 a.m.) > > > Review request for hive, Owen O'Malley and Gunther Hagleitner. > > > Description > ------- > > Hive CLI leaves behind the per session resource directory on non-interactive > invocation. The patch includes executing session state close() at the end of > non-interactive invocation. > Also changed the session id format to be a UUID. This is avoid possible > resource directory path conflict when there are multiple session HiveServer2 > from same user at same time. > > > This addresses bug HIVE-4546. > https://issues.apache.org/jira/browse/HIVE-4546 > > > Diffs > ----- > > cli/src/java/org/apache/hadoop/hive/cli/CliDriver.java 4239392 > ql/src/java/org/apache/hadoop/hive/ql/session/SessionState.java 8e6e24a > > Diff: https://reviews.apache.org/r/11083/diff/ > > > Testing > ------- > > > Thanks, > > Prasad Mujumdar > >
