Difference between running Postgres for yourself and for others

Re:Difference between running Postgres for yourself and for ... by solatic on Tuesday July 23, 2024 @02:05AM (#41042993)

"Point in time restore" is a bad way to call the feature if you don't let your customers pick a moment in time to restore to, so those tricks ensure that there's enough WAL entries to allow people to pick with per-second granularity.

[ Parent | on HN ]

Re:Difference between running Postgres for yourself and for ... by pintxo on Tuesday July 23, 2024 @02:47AM (#41043249)

One can let users pick a time and then just replay the closest backup? Why create empty backups?

[ Parent | on HN ]

Re:Difference between running Postgres for yourself and for ... by bjornsing on Tuesday July 23, 2024 @03:01AM (#41043345)

> those tricks ensure that there's enough WAL entries to allow people to pick with per-second granularity

How can they ensure that? Sounds more like they ensure that people can pick point in time with minute granularity.

I’m no expert, but doesn’t postgres write a timestamp with each record in the WAL? I know that when you start postgres in recovery mode it lets you specify a timestamp you want to restore to, and I don’t think that timestamp needs to be on a WAL file boundary. (So the tricks seem weird / unnecessary to me.)

[ Parent | on HN ]

Re:Difference between running Postgres for yourself and for ... by Doxin on Tuesday July 23, 2024 @03:00AM (#41043338)

But if there has been no activity you surely can just pick the most recent log that's older than the time the user picked?

[ Parent | on HN ]

Re:Difference between running Postgres for yourself and for ... by Tostino on Tuesday July 23, 2024 @08:42AM (#41045359)

How do you know the WAL was not lost during transmission, or the server crashed before transferring the WAL even though there were writes after the last WAL you have in your backup?

That's why.

[ Parent | on HN ]