|
Chris, [pg_dump et al] I think there's scope for creating some script that handles the backup automagically so it can be stuck in a cron job. The idea is to give people a framework to work with - adding FTP/SCP/DVD burning/switching on coffee makers is up to local conditions - and ideally this backup should be in a format that allows restore of the complete accounts as well as the code (also allows for making backups prior to fundamental structural changes in code/database - always nice to be able to go back the old "machine"). In addition to that, lesson no 1 of Business Continuity Processes is that a backup is not a backup unless you (a) can restore it and (b) regularly test and exercise the restore process - ergo a restore methodology would not go amiss either. Bottom line: good question that probably deserves more attention re answer, not everyone is a (very responsive) top coder like you :-). As for disasters, I've been there re. failing and incomplete backups - it's not the most popular discovery at 4am when you're busy reviving a system that is needed 9am.. This means that the steps prior to restoring the DB as you describe in part (2) of your answer probably need scripting too (most likely by using a complete tar.gz archive of code and db dump). /// P /// pg_dump is the supported way of doing backups in LedgerSMB. LedgerSMB uses pg_dump internally for backups. It is the best way to backup the database in this software and has the fewest problems. SQL-Ledger uses its own routines becase of the idea that some day non-PostgreSQL databases may again be supported. Also pg_dump prior to 8.0 had known issues handling dependencies and sometimes this meant that it was difficult or impossible to restore the data in a large backup. Since we only support PostgreSQL 8.0 and higher, these problems don't affect us. Also, just to help ensure that the process is understood, here is a little about the process: 1) pg_dump takes essentially a snapshot of the difference between template0 and the target database. This means that anything added to template1 or the current database is added. 2) To restore, you have to create a new database and pipe through psql. If you get errors, you might have something wrong with template1. You can fix this by creating your new database with the -T template0 option. Best Wishes, Chris Travers On 4/3/07, Joshua D. Drake <..hidden..> wrote: |