Connect PostgreSQL to Oracle 10g through ODBC: can't read all_tables

Question

Oracle 10g is not able to read the database table names (all_tables and all_objects, es. select * from all_tables@hspostgresodbc) from PostgreSQL using ODBC (v. 08.01.0200). This functionality is essential to be able to use Oracle Warehouse Builder and import external database definitions (heterogeneous services).
However, Oracle (SQL*Plus) is able to retrieve records when the table name is stated in the SQL query (es. select * from “my_table”).
The problem seems to be specifically with the PostgreSQL ODBC because Oracle is able to read schemes from other databases through ODBC. I’ve done some tests with Access dbs just to be sure that the problem is not with my system (Windows XP SP2).
The problem appears both with ANSI and UNICODE PostgreSQL drivers.
Any suggestion of how workaround this problem?
thanks in advance,
Raoni

Answer 1

Dear all,
I’ve tried different tracing levels with out success. The log file does not seem to be created by Oracle.
Looking closer to the debug and comm logs from the driver 8.01.0200, I’ve found out that the problem was generated after a table wasn’t found (the driver seems to search for the all_tables table, and just afterwards to query the scheme pg_catalog). In this case the all the following queries returned with the error “ERROR: current transaction is aborted, commands ignored until end of transaction block”. Maybe this problem is related with the bug [#1000641] http://pgfoundry.org/tracker/index.php?func=detail&aid=1000641&group_id=1000125&atid=538
Following the Postgres odbc developers suggestion, I’ve installed the psqlODBC v 08.02.0002 and I’ve got a completely different result.
As before, the simple query “select * from my_table@hsodbc” returns the result using SQL*Plus clearly faster than before.
Doing the query “select * from all_tables@hsodbc” the result is OK (the name of 56 tables) but the process is very slow and burdensome. The file paging memory in the machine running Oracle and the SQL*Plus jumps from the normal 700 MB to 1.7 GB, and the query takes about 20 minutes to complete.
Also the 1.7 GB remains allocated and the query process halted in PostGreSQL even after the on screen result of the query. The memory is freed and the process terminated after the logging out SQL*Plus.
Probably as result of this memory explosion, the system runs out of memory and Oracle Warehouse Builder selecting data base link -> schema, after some minutes get an error similar to: “can’t read from memory address XXXXX”
All those test where done with PostGreSQL 8.0.1 (on Linux Suse 9.0) and Oracle 10g (Windows XP SP2).
To be sure that the problem was not related to the PostGreSQL version, we updated the server to the 8.1.3 (running in Linux Suse 9.0 with psqlODBC v 08.02.0002 ) and OWB couldn’t see any table as before and doing Select * from all_tables@hsodbc we’ve got the message:
ERROR:
ORA-00604: error occurred at recursive SQL level 1
ORA-28500: connection from ORACLE to a non-Oracle system returned this message:
[Generic Connectivity Using ODBC]
ORA-02063: preceding 2 lines from HSPG
no rows selected
On the other hand, with the same configuration as before (Oracle 10g under Windows XP SP1, psqlODBC v 08.02.0002) but with PostGreSQL 8.1.3 under Windows XP SP1, OWB can read table names (same tables imported from Linux PostgreSQL) but not all_tables (a minor problem, because what is important for us is to make OWB work with PostgreSQL).
The encoding for all DBs is UTF8.
Any ideas of where could be the problem?
regards,
R.