OracleToMysql

A wrapper for the `sqlplus` and `mysql` binary that pulls data out of an Oracle database and inserts it into MySQL.

By omitting any Ruby/Mysql & Ruby/Oracle libraries things are kept

fast, minimal, and debuggable.

FYI, 1.0.0 is misleading, it's not some rock solid piece of uber-code. It's a weasel of a gem we just started using to replace our existing mirroring infrastructure in October 2010.

Use Case

As a Ruby developer with access to huge Oracle data warehouses I need a way to work with this data in my apps without using Oracle. ;)

System Dependencies

popen4 (gem) sqlplus, and mysql binaries are in $PATH

Usage

See the “test” cases in /test, there aren't really tests Just some code that uses the test/demo/* models to demonstrate the two main mirror strategies associated with this tool (see below)

For example: class TheMagicMirrorer include OracleToMysql

def otm_source_sql “select col || CHR(9) || col || CHR(9) || … from table where … Oracle statement” end

def otm_target_table “a_mysql_table_name” end def otm_target_sql “create table if not exists #selfself.otm_target_table (… mysql statement ” end end

x = TheMagicMirrorer.new
x.otm_execute

Will mirror the contents of otm_source_sql into the table created by otm_target_sql. The “|| CHR(9) ||” crap is Oracle sql code that tab deliminates the column content in the spooled sqlplus SQL data output. The Mysql “load data infile” command eats this output.

If you are using with Rails, it expects database.yml to have oracle_source and mysql_target entries to get the db connect info, to override the names, see OptionalOverrideInstanceMethods#otm_config_file and #otm_source_config_hash or #otm_target_config_hash

Also, will need to do:

gem install POpen4

Mirror Strategies

:atomic_rename (Default)
  "load data infile" the spooled oracle tab deliminted data into a temp table first
  then atomically rename 
    current_target_table -> old_target_table AND
    new_temp_table -> new_target_table

:accumulative
  "load data infile" directly into target_table replacing any existing
  rows in target when source data triggers "ON DUPLICATE KEY"

Target Table Retention

For both strategies, if the target_table already exists the default is to keep the existing table around and suffix it with “_old”. This is called a :n => 1 retention policy.

The ability to retain n previous tables in the works, handy for data mining, stay tuned.

Gem Development & Testing

The “tests” aka demo's assume you have a ps_term_tbl in your Oracle db. If you're running PeopleSoft at a University you'll probably have this… Otherwise, the tests won't run, it's just meant as an example that works in our world.

You'll need the thoughtbot-shoulda gem if you want to develop/hack on this gem or run the tests

To run tests: cd test

ruby test_oracle_to_mysql_against_ps_term_tbl.rb

OR

irb -r test_oracle_to_mysql_against_ps_term_tbl.rb
# And monkey with the run time...all files in test/demo are loaded, you can tinker with them

This assumes you have a connection file in the test dir:

oracle_to_mysql.yml

copy and populate from oracle_to_mysql.example.yml

Note on Patches/Pull Requests

  • Fork the project.

  • Add files to test/demo/* that demonstrate how you are using the tool.

  • Bugfixes = fork, fix, commit, pull request.

  • New Features = let us know what yer thinkin, we might already be working on it

A few things we'd like to work on soon

  • retention policy of 0: don't keep yesterdays data (aka don't create *_old table in mysql)

  • retention policy > 1: keep N mysql table copies around

  • usage of a Logger object instead of stdout

  • Better configuration of what happens when the otm_execute fails, not sure…some options might include

    • email someone a backtrace of the exception

    • log the exception backtrace to a table in the db

    • either cleanup + delete temp files or keep around (now it just leaves the temp files around)

Known Issues

  • Since source data is written to disk a tab delimintated file, if the source oracle data contains a t character in might mess things up (none of our data has tabs so we haven't had problems)

  • Add validations/checks for stuff in validate_otm_source_sql in write_sqlplus_commands_to_file.rb if you encounter goofy sqlplus errors Things that are not easily programmatically detectable ought just have an inline note i suppose