On Tue, Apr 24, 2018 at 10:50 AM, David Gauthier <davegauthierpg@xxxxxxxxx> wrote:
Typically, I would think doing a weekly full backup, daily incremental backups and turn on journaling to capture what goes on since the last backup.
This is almost the whole concept of the streaming replication built into postgres, except you are not applying the stream but archiving it. If you have atomic file system snapshots, you can implement this strategy along the lines of marking the DB snapshot for binary backup, snapshot the file system, then copy that snapshot file system off to another system (locally or off-site), meanwhile you accumulate the log files just as you would for streaming replication. Once the copy is done, you can release the file system snapshot and continue to archive the logs similarly to how you would send them to a remote system for being applied. You just don't apply them until you need to do the recovery.
Or just set up streaming replication to a hot-standby, because that's the right thing to do. For over a decade I did this with twin servers and slony1 replication. The cost of the duplicate hardware was nothing compared to not having downtime.
