Re: v12 pg_basebackup fails against older servers (take two) - Mailing list pgsql-hackers

From Stephen Frost
Subject Re: v12 pg_basebackup fails against older servers (take two)
Date
Msg-id 20191022130603.GY6962@tamriel.snowman.net
Whole thread Raw
In response to Re: v12 pg_basebackup fails against older servers (take two)  (Devrim Gündüz <devrim@gunduz.org>)
Responses Re: v12 pg_basebackup fails against older servers (take two)
List pgsql-hackers
Greetings,

* Devrim Gündüz (devrim@gunduz.org) wrote:
> On Tue, 2019-10-22 at 19:16 +0900, Michael Paquier wrote:
> > You are referring to the connection string generated in
> > primary_conninfo here, right?
>
> Right.

I'm awful suspicious that there's other similar cases beyond this
particular one...

> > It would be nice to be more compatible here.  This can be simply fixed by
> > having an extra filter in GenerateRecoveryConfig() (different file between
> > HEAD and REL_12_STABLE).  I also think that there is more.  On HEAD,
> > channel_binding gets added to the connection string generated which
> > would equally cause a failure with pg_basebackup from HEAD used for a
> > v12 or older server.
>
> https://git.postgresql.org/gitweb/?p=postgresql.git;a=commitdiff;h=beeb8e2e0717065296dc7b32daba2d66f0f931dd
>
> had a similar approach in backwards compatibility, so I also agree on fixing
> whatever breaks it.

Yeah, we clearly do want newer versions of pg_basebackup to work with
older versions of PG and therefore we should address this.

Here's just a quick rough-up of a patch (it compiles, I haven't tried it
out more than that) that adds in a check to skip gssencmode on older
versions.  If it seems like a reasonable approach then I can test it out
and deal with back-patching it and such.

Thoughts?

Thanks,

Stephen

Attachment

pgsql-hackers by date:

Previous
From: Thunder
Date:
Subject: Re:Re: [BUG] standby node can not provide service even it replaysall log files
Next
From: Michael Paquier
Date:
Subject: Re: v12 pg_basebackup fails against older servers (take two)