mirror of
https://git.postgresql.org/git/postgresql.git
synced 2024-12-09 08:10:09 +08:00
70c9763d48
change saves a great deal of space in pg_proc and its primary index, and it eliminates the former requirement that INDEX_MAX_KEYS and FUNC_MAX_ARGS have the same value. INDEX_MAX_KEYS is still embedded in the on-disk representation (because it affects index tuple header size), but FUNC_MAX_ARGS is not. I believe it would now be possible to increase FUNC_MAX_ARGS at little cost, but haven't experimented yet. There are still a lot of vestigial references to FUNC_MAX_ARGS, which I will clean up in a separate pass. However, getting rid of it altogether would require changing the FunctionCallInfoData struct, and I'm not sure I want to buy into that. |
||
---|---|---|
.. | ||
AddTrigger.sql | ||
clean_pending.pl | ||
DBMirror.pl | ||
Makefile | ||
MirrorSetup.sql | ||
pending.c | ||
README.dbmirror | ||
slaveDatabase.conf |
DBMirror - PostgreSQL Database Mirroring =================================================== DBMirror is a database mirroring system developed for the PostgreSQL database Written and maintained by Steven Singer(ssinger@navtechinc.com) (c) 2001-2004 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. Requirements: --------------------------------- -PostgreSQL-7.4 (Older versions are no longer supported) -Perl 5.6 or 5.8 (Other versions might work) -PgPerl (http://gborg.postgresql.org/project/pgperl/projdisplay.php) Upgrading from versions prior to 8.0 --------------------------------------- Users upgrading from a version of dbmirror prior to the one shipped with Postgresql 8.0 will need to perform the following steps 1. Dump the database then drop it (dropdb no not use the -C option) 2. Create database with createdb. 3. Run psql databasename -f MirrorSetup.sql 4. Restore the database(do not use the -C option of pg_dump/pg_restore) 5. run the SQL commands: DROP "Pending";DROP "PendingData"; DROP "MirrorHost"; DROP "MirroredTransaction"; The above steps are needed A) Because the names of the tables used by dbmirror to store data have changed and B) In order for sequences to be mirrored properly all serial types must be recreated. 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. PostgreSQL-8.0 Make Instructions: If you have already run "configure" in the top (pgsql) directory then run "make" in the dbmirror directory to compile the trigger. You should now have a file named pending.so that contains the trigger. Install this file in your Postgresql lib directory (/usr/local/pgsql/lib) 2) Run MirrorSetup.sql This file contains SQL commands to setup the Mirroring environment. This includes -Telling PostgreSQL about the "recordchange" trigger function. -Creating the dbmirror_Pending,dbmirror_PendingData,dbmirror_MirrorHost, dbmirror_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). 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 slaveName setting in the configuration file must match the slave name specified in the dbmirror_MirrorHost table. DBMirror.pl can be run in two modes of operation: A) It can connect directly to the slave database. To do this specify a slave database name and optional host and port along with a username and password. See slaveDatabase.conf for details. 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. B) The SQL statements that should be executed on the slave can be written to files which can then be executed slave database through psql. This would be suitable for setups where their is no direct connection between the slave database and the master. A file is generated for each transaction in the directory specified by TransactionFileDirectory. The file name contains the date/time the file was created along with the transaction id. 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" ("SlaveName") VALUES ('backup_system'); 6) Start DBMirror.pl DBMirror.pl is the perl script that handles the mirroring. It requires the Perl library Pg(See http://gborg.postgresql.org/project/pgperl/projdisplay.php) 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(not required if SQL files are being generated) 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 multi-master mirroring with conflict resolution. -Better support for dealing with Schema changes. Significant Changes Since 7.4 ---------------- -Support for mirroring SEQUENCE's -Support for unix domain sockets -Support for outputting slave SQL statements to a file -Changed the names of replication tables are now named dbmirror_pending etc.. Credits ----------- Achilleus Mantzios <achill@matrix.gatewaynet.com> Steven Singer Navtech Systems Support Inc. ssinger@navtechinc.com