On 2/19/18 2:05 PM, Don Seiler wrote: > On Mon, Feb 19, 2018 at 12:39 PM, David Steele <da...@pgmasters.net > <mailto:da...@pgmasters.net>> wrote: > > > I read "open it for testing (or backups in this case)" as letting > recovery complete and promoting the cluster to a master before taking > the backup. > > Don, is that the case? If it is, I think there's a problem with or > without a timeline switch. If you confirm the backup is being taken as > above then I'll detail my concerns. > > > Note that this is just for creating a couple of one-off backups to > restore for our dev and pre-prod environments. Given that, I was going > to open a new clone as its own cluster and take backups from that. The > data would be the same though and suit purposes of the dev and pre-prod > refreshes. > > If I were taking backups for the purpose of production backups, I would > not do things this way. That is the eventual plan but right now we > aren't ready to make the changes necessary in the production environment.
OK, that's fine then. You can play these to consistency and they'll be fine. I just wouldn't try to do any PITR using the production WAL archive. -- -David da...@pgmasters.net