On 2013-05-06 14:35:14 -0400, Stephen Frost wrote:
> * Andres Freund (andres@2ndquadrant.com) wrote:
> > Its rather useful if you e.g. want to instantiate a new replica without
> > rebuilding pg_dump/pg_restore's capabilities wrt. ordering, parallelism,
> > separating initial data load from index creation and all that. Which
> > already has been incompletely reinvented by several solutions :(.
>
> Perhaps it's just a wording thing, but I wouldn't use the term "replica"
> when referring to something built with pg_dump/restore- that should
> really be reserved for a slave system created through replication.
Well, part of the usecase *is* using it for the cloning in a replication
solution instead of open-coding it there. E.g. londiste and slony both
have home-grown hacks to do this.
> > So besides the above and real problems you point out this seems
> > worthwile to me...
>
> It certainly sounds interesting and I like the idea of it, but perhaps
> we need a different mechanism than just passing in a raw snapshot, to
> address the concerns that Tom raised.
If there is anything which isn't magnitudes more complex, I'd be
interested. But given we couldn't even find a sensible solution for
pg_dump internally I don't have all that high hopes...
Greetings,
Andres Freund
-- Andres Freund http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training &
Services