cdc_audit - Software to automate change data capture via triggers for mysql.

cdc_audit presently consists of two scripts:

• a script to auto generate mysql audit tables and triggers.
• a script to auto sync new rows in audit tables to a CSV file.

EXPERIMENTAL!

This software is still in an early prototype form. It has not yet been used on production systems and is not recommended for that purpose.

What problem does cdc_audit solve?

This software was written to solve a particular use-case, but may be useful for various change data capture tasks.

In our case terabytes of legacy data stored in multiple mysql databases. Users are demanding better and faster reporting and analytics capabilities than can be achieved with the transactional databases. Map-R has been installed and a fancy data analytics package has been chosen that enables end users to interact directly with the hadoop data and perform queries and visualizations as if they were working in a spreadsheet. However this data must regularly be synced from the mysql database that is collecting it to the Map-R filesystem. That's where cdc_audit can help.

We have a few requirements:

1. We need the sync process to be efficient. Only new/updated/deleted rows should be synced.

2. We prefer not to be modifying the source tables by hand, eg to add insert and update timestamps. We are looking for a tool that can automate this process for us.

3. We would prefer that the solution can accomodate deletes in addition to insert and update. Timestamping in the source table can't handle real deletes, only logical deletes, which requires custom logic in the application.

4. We have a goal not to require any modifications to the existing applications that use the source database tables.

5. We have some existing mysql triggers and cannot break or remove them.

How does cdc_audit meet these requirements?

cdc_audit breaks the CDC into three automated pieces:

a) creation of audit tables for every source table we are interested in. The audit tables contain all the columns of the source table plus audit columns: audit_event, audit_timestamp, and audit_pk. The audit_event column indicates the type of change to the source row: insert, update, or delete. The audit_pk column provides a unique id for each audit row, so we can easily identify new (unsynced) audit rows. audit_timestamp is unsuitable for this purpose because mysql only stores date/time columns to 1 second granularity.

b) creation of triggers on the source table(s) to insert rows into the audit table(s) whenever a row is inserted, updated, or deleted. This provides a complete audit history of all changes to the source table over time. Pre-existing trigger statements are included in the new trigger.

c) a script to sync new rows from the source audit table to a CSV file in the target (Map-R) filesystem.

The script cdc_audit_gen_mysql.php implements a) and b).

The script cdc_audit_sync_mysql.php implements c).

Features

• automates generation of audit tables, one audit table per source table.
• automates generation of triggers to populate audit tables
• automates syncing of new rows in audit tables to .csv files.
• Reads mysql information_schema to automatically determine tables and columns.
• fast triggers. only one insert and 0 selects per trigger execution.
• Can generate tables + triggers for all database tables, or a specified list.
• Can sync audit tables for all database tables, or a specified list.
• Retains pre-existing trigger logic, if any, when generating AFTER triggers.
• sync script option to delete all but last audit row, to keep source DB small.

Requirements

• PHP 5.3 or greater
• mysql 5.1 or greater

Usage

$ ./cdc_audit_gen_mysql.php

cdc_audit_gen_mysql.php [Options] -d [-h -d -u -p ]

Required: -d db mysql database name

Options:

-h HOST hostname of machine running mysql. default = localhost -u USER mysql username default = root -p PASS mysql password

-m audit_dir path to write db audit files. default = ./cdc_audit_gen.

-t tables comma separated list of tables. default = generate for all tables

-n namespace a prefix that will be pre-pended to all classnames. This makes it possible to use the generated classes multiple times in the same project. default = no prefix.

-o file Send all output to FILE -v Verbosity level. default = 1 0 = silent except fatal error. 1 = silent except warnings. 2 = informational 3 = debug. ( includes extra logging and source line numbers )

-?                Print this help.

$ ./cdc_audit_sync_mysql.php

cdc_audit_sync_mysql.php [Options] -d [-h -d -u -p ]

Required: -d db mysql database name

Options:

-h HOST hostname of machine running mysql. default = localhost -u USER mysql username default = root -p PASS mysql password

-m output_dir path to write db audit files. default = ./cdc_audit_sync.

-t tables comma separated list of tables. default = generate for all tables

-w wipe (incremental delete) all but the last audit row after syncing.

                  Note: this functionality is mostly untested!  dangerous!

-o file Send all output to FILE -v Verbosity level. default = 1 0 = silent except fatal error. 1 = silent except warnings. 2 = informational 3 = debug. ( includes extra logging and source line numbers )

-?                Print this help.

Usage Examples

First, unzip download package to a local directory anywhere.

To generate audit tables and triggers for all tables in a database:

php cdc_audit_gen_mysql.php -d <db> [-h <host> -d <db> -u <user> -p <pass>]

SQL file(s) will be generated in ./cdc_audit_gen. They can be applied to your database using the mysql command-line client, eg:

$ mysql -u root < ./cdc_audit_gen/table1.sql

To generate audit tables and triggers for a list of specific tables only:

php cdc_audit_gen_mysql.php -d <db> -t table1,table2,table3 [-h <host> -d <db> -u <user> -p <pass>]

To sync all audit tables in a database:

php cdc_audit_sync_mysql.php -d <db> [-h <host> -d <db> -u <user> -p <pass>]

To sync two specific audit tables in a database:

php cdc_audit_sync_mysql.php -d <db> -t table2_audit,table2_audit [-h <host> -d <db> -u <user> -p <pass>]

Once the sync process is running correctly, the command would typically be added to a unix crontab schduler in order to run it regularly.

Development

Development is taking place on github. Patches are welcome.

http://github.org/dan-da/cdc_audit

Known Issues

• If you make a change to the source table schema then the audit table and trigger will not reflect the change. You will need to alter the audit table manually then re-run cdc_audit_gen_mysql to recreate the triggers.

• no locking is performed on the target CSV file at present. This could cause file corruption.

Todos

• Use a lockfile to protect .CSV file. Map-R nfs does not support flock().

• Check the CSV header row when initiating sync to ensure that # of columns is unchanged and audit_pk column is correct.

• Auto-Detect schema changes to source table and apply to audit table.

• Enable creation of the audit tables in a separate database if desired.