Thread

Re: [HACKERS] Alternate locations for databases

Thomas Lockhart <lockhart@alumni.caltech.edu> — 1998-01-12T04:57:55Z

> Just checking to make sure this properly documented for 6.3 beta.  Is
> it?

man initlocationman create_database

Of course, it is not _properly_ documented. I haven't yet absorbed the man page
info into the new html/postscript docs, and may not be able to in time for v6.3.
Will see if anyone volunteers to help.

                                          - Tom

> > > >   CREATE DATABASE dbname WITH LOCATION = 'dbpath';
> > > How much work would it be to also be able to specify an alternate
> > > path for a table within a database?  I have some multi-Gb tables
> > > and am scrambling for room.  If I could have the tables on separate
> > > disks, that'd be wonderful.
> >
> > Well, it is (almost) trivial to get the full database in a different location;
> > in fact I put into service an unused column in pg_database which had clearly
> > been defined for this purpose. Probably not so trivial for individual tables,
> > indices, etc. If it is not on the ToDo list, perhaps Bruce could add it? I'm
> > probably not going to pursue it at the moment, myself, but would be happy to
> > work with someone if they want to do it :) As an aside, there is _no_
> > performance penalty for alternate database locations, but there might be for
> > distributed tables/indices since the location would need to be looked up at
> > least occasionally.