Developing SQL for automated execution

We have a team of developers who have very good knowledge of our applications. Sometimes the business wants the app changed but cannot make the changes through the application itself. The developers figure out which tables and columns need to be updated, test in non-production and then pass the script to the DBA team for execution. Often these scripts require inspecting the results of queries or other SQL statements and making a decision on whether to proceed with the next step or stop and roll back the changes. Not a big deal except when the business also requires the changes go into effect late at night or on the weekend.

I have been modifying the SQL for those off-hour execution times and then scheduling them to run at the expected time. I decided to train the development team so they could write the SQL themselves. They are new to Oracle but know SQL Server and SQL in general. I showed them how to write SQL that checks the actual results and rolls back or commits depending on expected results.

This works great and saves the DBA team a lot of late night work.

Here’s an example :

drop table SCHEMANAME.TABLENAME_BKP_changenum_B4;
 create table SCHEMANAME.TABLENAME_BKP_changenum_B4 as select * from SCHEMANAME.TABLENAME
 where
 (
 column1 in ('0040040','0040144','0040403','0040404','0040423','0040702')
 And column2 = '*ANY'
 ) or (
 column1 in ('0040040','0040144','0040403','0040404','0040423','0040702')
 ) or (
 column1 in ('0040019','0040085','0040086','0040145','0040405','0040406')
 And column2 <>'*ANY'
 ) or (
 column1 in ('54','55','56','57','59','79')
 And column2 <>'*ANY'
 );
 -- >>>> 210 ROWs or more is okay to proceed

declare
 exp_rows number;
 u_rows number;
 rowcount_fail exception;
 BEGIN
 dbms_session.set_identifier('changenumber');
 exp_rows:=41;
 dbms_output.put_line('Expecting '|| exp_rows || ' rows.');
 -- SQL#1 - DELETE #1

delete SCHEMANAME.TABLENAME
 WHERE
 (
 column1 in ('0040040','0040144','0040403','0040404','0040423','0040702')
 And column2 = '*ANY'
 );
 -- >>>> 41 ROWs or more is okay to proceed

u_rows:=SQL%ROWCOUNT;
 dbms_output.put_line('Updated ' || u_rows || ' rows.');
 if u_rows < exp_rows then
 raise rowcount_fail;
 else
 dbms_output.put_line('Rows updated >= expected row count');
 end if;

exp_rows:=27;
 dbms_output.put_line('Expecting '|| exp_rows || ' rows.');
 -- SQL #2 - update #1

UPDATE SCHEMANAME.TABLENAME
 set column3 = '*ANY',
 column3y = '*A',
 column4 = (100000 + to_char(sysdate,'YYDDD')),
 column5 = 'changenumber',
 colummn6 = SUBSTR(sys_context('USERENV','SERVER_HOST'),1,10)
 where
 (
 column1 in ('0040040','0040144','0040403','0040404','0040423','0040702')
 );
 -- >>>> 27 ROWs or more is okay to proceed

u_rows:=SQL%ROWCOUNT;
 dbms_output.put_line('Updated ' || u_rows || ' rows.');
 if u_rows < exp_rows then
 raise rowcount_fail;
 else
 dbms_output.put_line('Rows updated >= expected row count');
 end if;

exp_rows:=129;
 dbms_output.put_line('Expecting '|| exp_rows || ' rows.');
 -- SQL #3 - update #2

UPDATE SCHEMANAME.TABLENAME
 set column3 = '*ANY',
 column3y = '*A',
 column4 = (100000 + to_char(sysdate,'YYDDD')),
 column5 = 'changenumber',
 colummn6 = SUBSTR(sys_context('USERENV','SERVER_HOST'),1,10)
 where
 (
 column1 in ('0040019','0040085','0040086','0040145','0040405','0040406')
 And column2 <>'*ANY'
 );
 -- >>>> 129 ROWs or more is okay to proceed

u_rows:=SQL%ROWCOUNT;
 dbms_output.put_line('Updated ' || u_rows || ' rows.');
 if u_rows < exp_rows then
 raise rowcount_fail;
 else
 dbms_output.put_line('Rows updated >= expected row count');
 end if;

exp_rows:=13;
 dbms_output.put_line('Expecting '|| exp_rows || ' rows.');
 -- SQL#4 - DELETE #2

delete SCHEMANAME.TABLENAME
 WHERE
 (
 column1 in ('54','55','56','57','59','79')
 And column2 <>'*ANY'
 );
 -- >>>> 13 ROWs or more is okay to proceed

u_rows:=SQL%ROWCOUNT;
 dbms_output.put_line('Updated ' || u_rows || ' rows.');
 if u_rows < exp_rows then
 raise rowcount_fail;
 else
 dbms_output.put_line('Rows updated >= expected row count');
 end if;

--rollback;
 commit;

exception
 when rowcount_fail then
 dbms_output.put_line('Rows updated not >= expected row count, rolling back the change');
 rollback;
 when others then
 dbms_output.put_line('An error was encountered - '||SQLCODE||' -ERROR- '||SQLERRM);
 rollback;
 END;
 /

Advertisements

Setting up Ansible and cx_Oracle

I have been automating and centralizing dba and system administration processes using Python recently. One of the problems I solved was how to create Bash and SQL scripts to run on the remote database servers, copy them to the database servers and then execute them as another Linux account. After I got my Python scripts working perfectly, I discovered that I had re-invented Ansible 🙂

My scripts would have been much less complex and easier to write if I had done them with Ansible in mind. So that’s how I will write them going forward. I also wanted to implement a DBA Data Warehouse so I added cx_Oracle to the mix and used it to store the admin data in a database. I have been using a local virtual machine for Python development and the database was just a local installation on the VM.

Before moving to production I needed to figure out an easy way to get this stuff installed on the production Linux box. Setting it up on my VM was an adventure in Googling and trial and error. So here’s the documentation on how to set it up. You need to define the shell variables first.

export http_proxy=http://$my_userid:$my_password@$proxy_server:$port
export https_proxy=https://$my_userid:$my_password@$proxy_server:$port
wget https://bootstrap.pypa.io/ez_setup.py -O - | python
easy_install pip
export ORACLE_HOME=/base/oracle/product/12102
export LD_LIBRARY_PATH=$ORACLE_HOME/lib
yum install python-devel
pip install --proxy=https://$my_userid:$my_pasword@$proxy_server:$port cx_Oracle
pip install --proxy=https://$my_userid:$my_pasword@$proxy_server:$port ansible