On 2012-11-27 11:52:11 +0100, Andres Freund wrote:
> On 2012-11-27 11:48:08 +0100, Andres Freund wrote:
> > On 2012-11-27 10:18:37 +0530, Pavan Deolasee wrote:
> > > On Tue, Nov 27, 2012 at 9:01 AM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> > >
> > > > I wrote:
> > > >
> > > > Either state of indcheckxmin is valid with all three of these
> > > > combinations, so the specific kluge I was contemplating above doesn't
> > > > work. But there is no valid reason for an index to be in this state:
> > > >
> > > > indisvalid = true, indisready = false
> > > >
> > > > I suggest that to fix this for 9.2, we could abuse these flags by
> > > > defining that combination as meaning "ignore this index completely",
> > > > and having DROP INDEX CONCURRENTLY set this state during its final
> > > > wait before removing the index.
> > > >
> > > >
> > > Yeah, this looks much better, given our inability to add a new catalog
> > > column in 9.2. Can we cheat a little though and use a value other than 0
> > > and 1 for indisvalid or indisready to tell the server to interpret it
> > > differently ? I would assume not, but can't see a reason unless these
> > > values are converted to other types and back to boolean.
> > >
> > > Andres complained about the fact that many callers of RelationGetIndexList
> > > are probably not ready to process invalid or not-yet-ready indexes and
> > > suggested that those changes should be backpatched to even older releases.
> > > But IMO we should do that with a test case that demonstrates that there is
> > > indeed a bug.
> >
> > I haven't yet looked deeply enough to judge whether there are actually
> > bugs. But I can say that the e.g. the missing indisvalid checks in
> > transformFkeyCheckAttrs makes me pretty uneasy. Vacuum not checking
> > whether indexes are ready isn't nice either.
>
> At least the former was easy enough to verify after thinking about it
> for a minute (<=9.1):
>
> CREATE TABLE clusterbug(id serial primary key, data int);
> INSERT INTO clusterbug(data) VALUES(2),(2);
> CREATE UNIQUE INDEX CONCURRENTLY clusterbug_data ON clusterbug(data);
> CREATE TABLE clusterbug_ref(id serial primary key, clusterbug_id int references clusterbug(data));
>
> Now a !indisready index is getting queried (strangely enough that
> doesn't cause an error).
That should work in 9.2 as well, although its slightly more
complicated. You just need a indisready && !indisvalid index and the
foreign key will happily be created. Easy enough to achieve with two
sessions.
Greetings,
Andres Freund
--Andres Freund http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training &
Services
