postgres backup pgbackrest - ghdrako/doc_snipets GitHub Wiki
- ma własny protokół transferu plików - wymaga instancji agenta na serwerach
- odtworzenia się z katalogu - czyli uruchomieniu bazy z katalogu backupu na szybko
- backup ze standby (wal-e z mastera)
- backup różnicowy na poziomie blocku
It is built by the CrunchyData company and written in Perl. Features of this tool are:
- Backup scheduling
- Log archiving
- Retention management
- Multiserver
- Restore shortcuts
- Obsessive validation
It’s MIT licensed. It uses SSH but it doesn’t use rsync. It uses its own protocol tunneled over SSH. The protocol is the enabler of the features that pgBackRest has that Barman doesn’t. In particular, it supports parallel backup sessions so you can scale out and make your backups run faster. For most people, single threaded backups are not really a problem. But if your database is big, being able to do multithreaded backups can save your backup time significantly and also obviously your restore time when you’re getting things back, which is more important. It also supports full differential and incremental backups, and it does this at a segment basis. In PostgreSQL, we have all our tables split into segments of one gigabyte, and backrest basically functions like “if nothing in that one gigabyte has been modified, then I don’t need to back it up again. If even a single byte has been modified, I’ll copy the whole segment.” Because if you actually want to look at every block, it’ll just take too long to figure out if something has changed. Doing it at a gigabyte means if large portions of your database are read-only, your backups will be much faster because we can just skip that and get it from the previous full or differential backups. It does not support concurrent backups