Re: JDBC interpreter is not concurrent

[Boris Shminke]

Hi! This PR https://issues.apache.org/jira/browse/ZEPPELIN-995 helped a bit but still the JDBC interpreter is somehow different. First, it shows very weird error messages (actually, pretty standard Java error stack but they are not about SQL and I can't understand what's wrong with my queries). Hive interpreter, on the contrary, threw absolutely normal SQL error messages which helped in diagnosing the problem. Another thing is related to so called local temporary tables in HP Vertica (I use JDBC interpreter to send queries to it). I don't know how it relates to Hive or JDBC interpreter in Zeppelin but when I create such a table in one cell of the notebook I can not access it in another. I guess that means that JDBC interpreter creates separate connections to DBMS in separate cell. That was not the case with Hive interpreter, was it? I like the idea of JDBC interpreter but why it differs so much from old Hive interpreter? On Mon, 13 Jun 2016 at 13:56 Jongyoul Lee <jongy...@gmail.com> wrote: I made a PR for dealing with it.. Check https://issues.apache.org/jira/browse/ZEPPELIN-995 @Prasad Concerning maxConcurrency, I'm trying to change Zeppelin's structure in order to support it. Wait a moment please. JL On Sun, Jun 12, 2016 at 2:29 AM, Jongyoul Lee <jongy...@gmail.com> wrote: Hello, Thanks for reporting it. I will make a PR for solving it. Regards, JL On Saturday, 11 June 2016, Timur Shenkao <t...@timshenkao.su> wrote: Hi guys! I agree with Boris that this feature (ability to issue multiple Hive queries simultaneously) is extremely important as people use Hive interpereter not only against Hive but against other systems, too. Predominantly RDBMS. Making changes in source code by ourselves (not to mention without tests) leads to incomatibility with future releases and doesnt' solve problem. On Fri, Jun 10, 2016 at 10:22 PM, Prasad Wagle <prasadwa...@gmail.com> wrote: Hi Boris, You can run concurrent queries by overriding the getScheduler function in JDBCInterpreter to return ParallelScheduler like HiveInterpreter. @jongyoul - should we change JDBCInterpreter. getScheduler to return ParallelScheduler? It would also be nice to make maxConcurrency configurable. JDBCInterpreter:   @Override   public Scheduler getScheduler() {     return SchedulerFactory.singleton().createOrGetFIFOScheduler(         JDBCInterpreter.class.getName() + this.hashCode());   } HiveInterpreter:   @Override   public Scheduler getScheduler() {     return SchedulerFactory.singleton().createOrGetParallelScheduler(         HiveInterpreter.class.getName() + this.hashCode(), 10);   } Thanks, Prasad On Fri, Jun 10, 2016 at 5:52 AM, Boris Schminke <schmink...@gmail.com> wrote: Hi all, when I used the Hive interpreter I could send several queries in parallel. Now in Zeppelin 0.6 Hive Interpreter is merged into JDBC (which is very good) but I can not execute concurrent Hive queries (which is not so good). How can I run concurrent queries to Hive using JDBC interpreter now in Zeppelin 0.6? Regards, Boris -- 이종열, Jongyoul Lee, 李宗烈 http://madeng.net -- 이종열, Jongyoul Lee, 李宗烈 http://madeng.net