On Tue, 5 Mar 2019 at 18:39, Adrian Klaver <adrian.klaver@xxxxxxxxxxx> wrote:
>
> The replication documentation, and more specifically the pg_basebackup
> documentation, makes no mention of cross-OS replication as being a
> problem for any reason. If that is expected to be a problem, then
> perhaps that should be updated?
Generally covered under:
https://www.postgresql.org/docs/10/warm-standby.html#STANDBY-PLANNING
"It is usually wise to create the primary and standby servers so that
they are as similar as possible, at least from the perspective of the
database server."
Nothing in that paragraph says to me that I'm going to have problems as a result of differences in postgres software dependencies. It's a pretty vague warning that seems to imply that as long as your hardware architecture and filesystem layout are identical there shouldn't be any issues.
Thanks for your clarification, though. We'll have to take that into account in our migration plan back to the original server. I don't think we can afford the downtime for a dump/restore between systems, so we'll likely just end up regenerating indexes again, before cutting the application over to the other DB server.
That is why later versions(10+) grew logical replication:
https://www.postgresql.org/docs/10/logical-replication.html
That seems promising, but we'd have to deal with the upgrade to 10.x first. And presumably logical replication would allow more freedom in replicating between mismatched versions as well... possibly as a low-downtime migration path for getting very large databases moved to more recent releases?
