Re: Logical replication and multimaster

On 2015-12-11 19:24, Robert Haas wrote:
> On Fri, Dec 11, 2015 at 5:16 AM, Andres Freund <andres(at)anarazel(dot)de> wrote:
>> On 2015-12-11 18:12:55 +0800, Craig Ringer wrote:
>>> On 10 December 2015 at 03:19, Robert Haas <robertmhaas(at)gmail(dot)com> wrote:
>>>> On Sun, Dec 6, 2015 at 10:24 PM, Craig Ringer <craig(at)2ndquadrant(dot)com>
>>>> wrote:
>>>>> * A way to securely make a libpq connection from a bgworker without
>>>> messing
>>>>> with passwords etc. Generate one-time cookies, sometihng like that.
>>>>
>>>> Why would you have the bgworker connect to the database via TCP
>>>> instead of just doing whatever it wants to do directly?
>>
>>> pg_dump and pg_restore, mainly, for copying the initial database state.
>>
>> Well, you don't want to necessarily directly connect from the bgworker,
>> but from processes started from a bgworker. I guess that's where a good
>> bit of the Robert's confusion originated.
>
> That's part of it, yeah. I'm a little scared of this design. I mean,
> I understand now why Craig wants to do this (thanks for explaining,
> Craig!), but it seems like it's going to have a lot of the same
> reliability problems that pg_upgrade does. I'm not saying there's a
> better way to get the functionality, but it's pretty obvious that
> depending on tools other than the server itself, and in particular
> pg_dump, vastly increases the failure surface area.
>

Well, it's better than trying to write completely new catalogs dump tool
for this. As Craig said, it would be best if pg_dump functionality was
moved to functions in a backend, but that's probably not gonna happen
tomorrow.

--
Petr Jelinek http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services