Re: Re: BUG #10329: Could not read block 0 in file "base/56100265/57047884": read only 0 of 8192 bytes

On Thu, Sep 11, 2014 at 07:29:23PM +0200, Andres Freund wrote:
> On 2014-09-11 19:25:13 +0200, Stefan Kaltenbrunner wrote:
> > On 09/08/2014 03:45 PM, Bruce Momjian wrote:
> > > On Sat, Sep  6, 2014 at 09:42:45PM -0400, Alvaro Herrera wrote:
> > >> Bruce Momjian wrote:
> > >>
> > >>> Here is a patch which implements the warning during CREATE INDEX ...
> > >>> HASH.  If WAL-logging of hash indexes is ever implemented, we can remove
> > >>> this warning.
> > >>
> > >> I think we should have CREATE UNLOGGED INDEX, and simply disallow any
> > >> hash index from being created unless it's marked as such.
> > >
> > > Wow, that sounds much more radical than we discussed.  Seeing I got
> > > push-back just for the warning, I don't see how disabling "logged" WAL
> > > indexes is going to be accepted.
> > >
> > > It is a good idea, though.  :-)
> >
> > I agree there - implementing CREATE UNLOGGED INDEX and use THAT for hash
> > indexes seems like a fairly clean thing to me, hash indexes _are_
> > unlogged so lets reflect that directly.
> > I could even envision pg_dump doing that conversion automatically...
>
> I think this did came up as a solution before. It's just that nobody
> found a reasonably easy and clean way to do unlogged indexes on logged
> tables so far. It's far from trivial.

And practically, how would we implement this for upgrades?  Would we have
pg_dump emit UNLOGGED for any hash creation command?  That seems to
defeat the purpose of this.

--
  Bruce Momjian  <bruce@momjian.us>        http://momjian.us
  EnterpriseDB                             http://enterprisedb.com

  + Everyone has their own god. +