Re: Using a single standalone-backend run in initdb (was Re: Bootstrap DATA is a pita)

Andres Freund <andres@anarazel.de> writes:
> On 2015-12-12 17:31:49 -0500, Tom Lane wrote:
>> Does anyone know of people using standalone mode other than
>> for initdb?

> Unfortunately yes. There's docker instances around that configure users
> and everything using it.

Hm, that means that we *do* have to worry about backwards compatibility.
We might be able to get away with changing the behavior of -j mode anyway,
though, since this proposal mostly only changes when execution happens
and not what is valid input for -j mode.  (It would probably break some
apps if we took away the switch, since right now, you do not need a
semicolon to terminate commands in the regular standalone mode.)
Failing that, we could define a new switch, I guess.

> That's cool too. Besides processing the .bki files, and there largely
> reg*_in, the many restarts are the most expensive parts of initdb.

Right.  The proposal we were discussing upthread would move all the reg*
lookups into creation of the .bki file, basically, which would improve
that part of things quite a bit.  (BTW, if we are concerned about initdb
speed, that might be a reason not to be too eager to shift processing
from bootstrap to non-bootstrap mode.  Other than the reg* issue,
bootstrap is certainly a far faster way to put rows into the catalogs
than individual SQL commands could be.)
        regards, tom lane