On 23 August 2010 09:26, Dave Kempe <[email protected]> wrote: > ----- Original Message ----- >> From: "james" <[email protected]> >> So the file created on day 2 backed up on day 3 is lost. >> Can anybody point to the boat (I've missed) or confirm my vision. > > Seems correct, and the primary reason why we moved to rdiff-backup and disk > based backup. > We get to keep effective full versions for as long as we have disk space, > very efficiently. We have some customers have 500 days worth at least of > their core database dumps. > They needed to restore something back to 400 or so days once, and where able > to do it in a matter of minutes, from request to delivery of the file. > It works for us, and we have at least 10Tb of backup jobs under management > spread over hundreds of backup jobs. > > I also agree with another poster - the typical restore event is the most > recent backup, and a subset of the job at that. > > http://rdiff-backup.nongnu.org/ > and > https://labs.riseup.net/code/projects/show/backupninja > to manage it. (Well a forked version of it with a few extra features...)
How do you backup databases with that (I'm specifically interested to hear about PostgresQL)? Snapshot the filesystem every day? We have a 300Gb postgres database and growing rapidly, and I'm looking for better ways to back it up with point in time recovery other than backing up the transaction logs and periodic full backups. We also have around similar size of data spread across hundreds of sqlite files. For now we take Logical Volume snapshots then backup the file from them. We use Bacula to do that. Thanks, --Amos -- SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/ Subscription info and FAQs: http://slug.org.au/faq/mailinglists.html
