I ran into a odd situation yesterday.  I got handed a performance problem that had been going on for months, the developers were blaming the database people, the database people were producing useless reports and statistics and around and around it went.  All the end user wanted to do was view a journal entry in the finance system from the web tier.

Well, my initial thought most likely the journal line table was really large and there may be need for another index.  The journal line table had 1.1 million rows, but the row count SQL (select count(1) from ps_jrnl_ln) was taking a really long time to return, when I selected the data by the first two key fields to return a specific journal (select count(1) from ps_jrnl_ln where business_unit = ‘SHARE’ and journal_id = ‘PAY1234’).  The RED flag is waving madly at this point. So here is what I did:

1) Opened the Record: JRNL_LN in application designer and did an alter table build (w/even if no change flag on).  The build produced the following set of errors:

ORA-12801: error signaled in parallel query server xyz
ORA-01452: cannot CREATE UNIQUE INDEX; duplicate keys found

How in the world did that happen?  Good question, can this really happen?  Not under normal operation, but it has happened so how do we fix it:

select business_unit, journal_id, journal_date, unpost_seq, journal_line, ledger, count(1)
from ps_jrnl_ln
group by business_unit, journal_id, journal_date, unpost_seq, journal_line, ledger
having count(1) > 1;


This identified 31 rows having a count greater than 1.  So I looked at a couple and found they were completely identical rows. So, how to get rid of them.  I used the returned data from the previous select and built and export script in datamover that basically exported all the duplicate rows to a data file.

set log l:\data\logs\export_bad_data.log;
set output l:\data\bad_data.dat;
export jrnl_ln where business_unit = 'xyz' and journal_id = '123';
rem ..... (fill in all the key fields from the above select);
export jrnl_ln where business_unit = 'xyz' and journal_id = '128'
rem ..... so on until all the rows are exported;


Now, I exited datamover, and went back into datamover in bootstrap mode. I imported the data using ignore_dups and the as command, to import the data into a temporary table.

set log l:\data\logs\import_bad_data.log;
set input l:\data\bad_data.dat;
set ignore_dups;
import jrnl_ln as temp_jrnl_ln;

Now, I have the 31 unique rows in a temporary table.  I delete all the duplicate rows from the jrnl_ln table, using the results from the first select, I turn the results into delete statements by all the keys.  This deletes a total of 62 rows from the jrnl_ln table. I then do a insert from the temp table, which puts the 31 unique rows back into the jrnl_ln table, and I drop the temp table.

Rebuild the index, and it returns a success with no warnings or errors.  On the web, I open up the journal in question in a couple of seconds.   !!!!!!!!!!!

When working in one database you can make reference to data in another database by creating a database link.

To create an Oracle database link, the quickest way is to just create a public database link using a user id in the other database that has permissions to reference the data you need. {LINKNAME} = the name for the link you want to create, {LINKUSER} = the user on the remote database that has access, and {LINKUSERPASSWORD} is that users password.  {TNSNAMES-DB-ENTRY} is the database entry name from the TNSNAMES.ora file.

create public database link {LINKNAME} connect to {LINKUSER} identified by {LINKUSERPASSWORD using ‘{TNSNAMES-DB-ENTRY}’;

To Reference the data from the linked database, simply add the link identifier to the table you are selecting – for example:

select * from remotetable@{LINKNAME};

To Remove the database link:

drop public database link {LINKNAME};

In PeopleSoft you can use datamover to add SQL security.  The reason for this is even though you log into PeopleSoft as a specific user once the user is verified it actually uses the user that is associated with the psaccessprfl access profile.  This means that in oracle typically the user sysadm is the user behind the access profile, and since the sysadm user is typically built using the PSADMIN role profile, you have priviledges to run a lot of commands from datamover.

A good example of this is when you build a table in PeopleSoft but you can’t see it in SQL because the user account you have in Oracle does not access to the new table.  So in order to grant access to this you can do a couple of different statements.  The simplest is to grant select access on the new table to a specific SQL user account:

grant select on new_table to sql_user;

Often times there are roles setup that are assigned to a user, if you do a:

select * from dba_roles;

You can see all the roles that are on the system.  Typically there is a role that will be setup that will allow “select” access and another role that is setup to allow “select, update, delete, insert”, sometimes referred to as SUDI access.  So lets say for example there is a role called SUDI_ALL_TABS, you can issue the statement in datamover:

grant select, update, delete, insert on new_table to SUDI_ALL_TABS;

Note: The new_table needs to have the ps_ qualifier on it, as this is not a typical function of datamover and therefore needs the fully qualified table name.

Outer joins is just one of the items that is required from time to time, but since I typically never recommend using them with PeopleSoft, I always forget the syntax.  Since my days go back to Oracle 7, I remember the messy (+) notation and so I always end up looking it up.  The good news is Oracle superseded the (+) notation by going to ISO 99 standard as of Oracle 9i, however it still supports (+) notations.

There are three types of joins you can do, left outer, right outer and full outer join. In a left outer join, all the records of the table listed on the left will be returned with the combined results of the join to the table listed on the right, however, if the right listed table does not have a matching row, NULLS will be returned for the columns of the right listed table.  The right outer join does the same thing just in reverse, whereas the full outer join basically does a UNION of the left & right outer joins.

Some database platforms do not support full outer joins so you can basically do it by doing a left outer join combined with with a right outer join using the union statement.

SELECT A.NAME, D.DEPARTMENT FROM EMPLOYEE A LEFT OUTER JOIN DEPARTMENT D ON A.DEPTID = D.DEPTID;

If you want to do the Oracle (+) syntax:

SELECT A.NAME, D.DEPARTMENT FROM EMPLOYEE A, DEPARTMENT D WHERE A.DEPTID = E.DEPTID (+);

When you are at that client site that insists that you must use your own account to interact with PeopleSoft but they don’t set your schema, just change your default schema using the command:

alter session set current_schema=sysadm;

Where sysadm is the default PeopleSoft schema in Oracle, however, PeopleSoft does not have any issues with the schema owner being different then sysadm.  A couple of ways to find the schema, is to look at the table:  psdbowner (ownerid) or psstatus (ownerid). By default the psdbowner table is stored in the PS schema.