postgresql/contrib/dbmirror/README.dbmirror

DBMirror - Postgres Database Mirroring
===================================================


DBMirror is a database mirroring system developed for the Postgres
database Written and maintained by Steven Singer(ssinger@navtechinc.com)


(c) 2001-2002 Navtech Systems Support Inc.
ALL RIGHTS RESERVED

  Permission to use, copy, modify, and distribute this software and its
  documentation for any purpose, without fee, and without a written agreement
  is hereby granted, provided that the above copyright notice and this
  paragraph and the following two paragraphs appear in all copies.
 
  IN NO EVENT SHALL THE AUTHOR OR DISTRIBUTORS BE LIABLE TO ANY PARTY FOR
  DIRECT, INDIRECT, SPECIAL, INCIDENTAL, OR CONSEQUENTIAL DAMAGES, INCLUDING
  LOST PROFITS, ARISING OUT OF THE USE OF THIS SOFTWARE AND ITS
  DOCUMENTATION, EVEN IF THE AUTHOR OR DISTRIBUTORS HAVE BEEN ADVISED OF THE
  POSSIBILITY OF SUCH DAMAGE.
 
  THE AUTHOR AND DISTRIBUTORS SPECIFICALLY DISCLAIMS ANY WARRANTIES,
  INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY
  AND FITNESS FOR A PARTICULAR PURPOSE.  THE SOFTWARE PROVIDED HEREUNDER IS
  ON AN "AS IS" BASIS, AND THE AUTHOR AND DISTRIBUTORS HAS NO OBLIGATIONS TO
  PROVIDE MAINTENANCE, SUPPORT, UPDATES, ENHANCEMENTS, OR MODIFICATIONS.



Overview
--------------------------------------------------------------------

The mirroring system is trigger based and provides the following key features:

-Support for multiple mirror slaves
-Transactions are maintained
-Per table selection of what gets mirrored.


The system is based on the idea that a master database exists where all
edits are made to the tables being mirrored.   A trigger attached to the
tables being mirrored runs logging information about the edit to 
the Pending table and  PendingData table. 

A perl script(DBMirror.pl) runs continuously for each slave database(A database
that the change is supposed to be mirrored to) examining the Pending
table; searching for transactions that need to be sent to that particular slave 
database.  Those transactions are then mirrored to the slave database and
the MirroredTransaction table is updated to reflect that the transaction has
been sent.

If the transaction has been sent to all know slave hosts (All entries
in the MirrorHost table) then all records of it are purged from the
Pending tables.

Installation Instructions
------------------------------------------------------------------------


1) Compile pending.c

The file pending.c contains the recordchange trigger.  This runs every
time a row inside of a table being mirrored changes.


To build the trigger run  make on the "Makefile" in the DBMirror directory.

Postgres-7.3 Make Instructions:

  If you have already run "configure" in the pgsql-server directory 
  then run "make" in the dbmirror directory to compile the trigger.

Postgres-7.1 & Postgres-7.2 Make Instructions:

  The included Makefile is not compatible with postgres 7.1 and 7.2
  The trigger will need to be built by hand.

  Run the following commands

  gcc -fpic -I/usr/local/pgsql/include/server -c pending.c -DNOSCHEMAS
  ld -shared -o pending.so pending.o

  Assuming the postgres include files are in /usr/local/pgsql/include/server.  

Postgres-7.1.x installations should change this to
/usr/local/pgsql/include  (The server part is for 7.2+)

If you have installed the postgres include files to another location then
  modify the include path to reflect this.

  Compiling the trigger by hand requires that all postgres headers be installed
  ,this is accomplished in postgresql(7.1 or 7.2) by running 
  "make install-all-headers" in the postgres source directory.


You should now have a file named pending.so that contains the trigger.

Install this file in /usr/local/pgsql/lib  (or another suitable location).

If you choose a different location the MirrorSetup.sql script will need
to be modified to reflect your new location.   The CREATE FUNCTION command
in the MirrorSetup.sql script associates the trigger function with the
pending.so shared library.  Modify the arguments to this command if you
choose to install the trigger elsewhere.

2) Run MirrorSetup.sql

This file contains SQL commands to setup the Mirroring environment.  
This includes

-Telling Postgres about the "recordchange" trigger function.
-Creating the Pending,PendingData, MirrorHost, MirroredTransaction tables


To execute the script use psql as follows 

"psql -f MirrorSetup.sql  MyDatabaseName"

where MyDatabaseName is the name of the database you wish to install mirroring
on(Your master).

Postgres-7.1 and 7.2 Notes:
  -The syntax for creating a trigger function changed in Postgres-7.3. 
  Change the line in MirrorSetup.sql from 

  CREATE FUNCTION "recordchange" () RETURNS trigger AS

  to 
  CREATE FUNCTION "recordchange" () RETURNS OPAQUE AS


3) Create slaveDatabase.conf files.

Each slave database needs its own configuration file for the 
DBMirror.pl script.  See slaveDatabase.conf for a sample.

The master settings refer to the master database(The one that is
being mirrored).

The slave settings refer to the database that the data is being mirrored to.
The slaveHost parameter must refer to the machine name of the slave (Either
a resolvable hostname or an IP address).  The value for slave host
must match the Hostname field in the MirrorHost table(See step 6).

The master user must have sufficient permissions to modify the Pending
tables and to read all of the tables being mirrored.

The slave user must have enough permissions on the slave database to
modify(INSERT,UPDATE,DELETE) any tables on the slave system that are being
mirrored. 

4) Add the trigger to tables.

Execute the SQL code in AddTrigger.sql once for each table that should
be mirrored.   Replace MyTableName with the name of the table that should
be mirrored.

NOTE: DBMirror requires that every table being mirrored have a primary key
defined.

5)  Create the slave database.

The DBMirror system keeps the contents of mirrored tables identical on the
master and slave databases.  When you first install the mirror triggers the
master and slave databases must be the same.

If you are starting with an empty master database then the slave should
be empty as well.  Otherwise use pg_dump to ensure that the slave database
tables are initially identical to the master.

6) Add entries in the MirrorHost table.

Each slave database must have an entry in the MirrorHost table.

The name of the host in the MirrorHost table must exactly match the
slaveHost variable for that slave in the configuration file.

For example
INSERT INTO "MirrorHost" ("HostName") VALUES ('mySlaveMachine.mycompany.com');


6)  Start DBMirror.pl


DBMirror.pl is the perl script that handles the mirroring.  

It requires the Perl library Pg(See src/interfaces/perl5 in the postgres 
source distribution).

It takes its configuration file as an argument(The one from step 3)
One instance of DBMirror.pl runs for each slave machine that is receiving
mirrored data.

Any errors are printed to standard out and emailed to the address specified in
the configuration file. 

DBMirror can be run from the master, the slave, or a third machine as long
as it is able to access both the master and slave databases.

7) Periodically run clean_pending.pl 
clean_pending.pl cleans out any entries from the Pending tables that
have already been mirrored to all hosts in the MirrorHost table.
It uses the same configuration file as DBMirror.pl.

Normally DBMirror.pl will clean these tables as it goes but in some 
circumstances this will not happen.

For example if a transaction has been mirrored to all slaves except for
one, then that host is removed from the MirrorHost table(It stops being
a mirror slave) the transactions that had already been mirrored to 
all the other hosts will not be deleted from the Pending tables by 
DBMirror.pl since DBMirror.pl will run against these transactions again
since they have already been sent to all the other hosts.

clean_pending.pl will remove these transactions.

TODO(Current Limitations)
----------
-Support for selective mirroring based on the content of data.
-Support for BLOB's.
-Support for conflict resolution.
-Batching SQL commands in DBMirror for better performance over WAN's.
-Better support for dealing with Schema changes.

Tested Platforms:
------------------

DBMirror has been tested on the following configurations but should
work on any platform with Postgres >= 7.1 and Perl 5.6. 

RedHat Linux 7.1 & 6.2
 -Postgres 7.1.2
 -Perl 5.6

Mandrake Linux 8.0(Limited Testing)
 -Postgres 7.2
 -Postgres 7.3
 -Perl 5.6


Steven Singer
Navtech Systems Support Inc.
ssinger@navtechinc.com