[jira] [Created] (HIVE-14804) HPLSQL multiple db connection does not switch back to Hive

["Dmitry Kozlov (JIRA)" <ji...@apache.org>]

Dmitry Kozlov created HIVE-14804:
------------------------------------

             Summary: HPLSQL multiple db connection does not switch back to Hive
                 Key: HIVE-14804
                 URL: https://issues.apache.org/jira/browse/HIVE-14804
             Project: Hive
          Issue Type: Bug
          Components: hpl/sql
            Reporter: Dmitry Kozlov
            Assignee: Dmitry Tolpeko
            Priority: Blocker


I have a problem with a multi database connection. I have 3 environments that I would like to connect in my HPLSQL code Hive, DB2 and MySql. As soon as I map any table either from DB2 or MySQL my code stops to recognize Hive tables. Actually it starts to think that it is a table from the same database (DB2 or MySql) that was mapped the last. It means your example http://www.hplsql.org/map-object works only one way from Hive to MySQL and it is not possible to go back to Hive.  


Here is a simple piece of code.

declare cnt int;
begin

/*
PRINT 'Start MySQL';
MAP OBJECT tbls TO hive.TBLS AT mysqlconn;
select count(*)
into cnt
from tbls;
PRINT cnt;

PRINT 'Start Db2';

MAP OBJECT exch TO DBDEV2.TEST_EXCHANGE AT db2conn;
select count(1) 
into cnt
from exch;
PRINT cnt;*/

PRINT 'Check Hive';
SELECT count(1) 
into cnt
FROM dev.test_sqoop;
PRINT cnt;

end;

It has three blocks. One select from MySQL, second from DB2 and third from Hive ORC table.
When first two blocks are commented then block 3 works. See below
Check Hive
16/09/20 18:08:08 INFO jdbc.Utils: Supplied authorities: localhost:10000
16/09/20 18:08:08 INFO jdbc.Utils: Resolved authority: localhost:10000
16/09/20 18:08:08 INFO jdbc.HiveConnection: Will try to open client transport with JDBC Uri: jdbc:hive2://localhost:10000
Open connection: jdbc:hive2://localhost:10000 (497 ms)
Starting query
Query executed successfully (177 ms)
82

When I try to uncomment any of those blocks then block 3 stops working. For example, if I uncomment block 1 I get this output. It is now assumes that dev.test_sqoop is a MySQL table. Contrarily to your example

Start MySQL
Open connection: jdbc:mysql://10.11.12.144:3306/hive (489 ms)
Starting query
Query executed successfully (4 ms)
539
Check Hive
Starting query
Unhandled exception in HPL/SQL
com.mysql.jdbc.exceptions.jdbc4.MySQLSyntaxErrorException: Table 'dev.test_sqoop' doesn't exist

If I comment the second block then it starts to assume that dev.test_sqoop is a DB2 table. See below. So switch between DB2 and MySQL is working, however, the hive table is still not working

Start MySQL
Open connection: jdbc:mysql://10.11.12.144:3306/hive (485 ms)
Starting query
Query executed successfully (5 ms)
539
Start Db2
Open connection: jdbc:db2://10.11.12.141:50000/WM (227 ms)
Starting query
Query executed successfully (48 ms)
0
Check Hive
Starting query
Unhandled exception in HPL/SQL
com.ibm.db2.jcc.am.SqlSyntaxErrorException: DB2 SQL Error: SQLCODE=-204, SQLSTATE=42704, SQLERRMC=DEV.TEST_SQOOP, DRIVER=4.16.53

Could you, please, provide your feedback on this finding. In addition, I would like to check if it would be possible to insert into a DB2 table records that were selected from a Hive with one statement as soon as DB2 table is properly mapped. Please, explain.

Looking forward to hearing from you soon.


Regards,
Dmitry Kozlov
Daisy Intelligence   




--
This message was sent by Atlassian JIRA
(v6.3.4#6332)