Re: Question: pg_class attributes and race conditions ?

Tom Lane wrote:
> "Pavan Deolasee" <pavan(dot)deolasee(at)enterprisedb(dot)com> writes:
>> Any thoughts on the overall approach ?
>
> Fragile and full of race conditions :-(.
>

Yes, it looks a bit complex. But IMHO we can get around that.
Do you have any ideas in mind about doing that ?

> I thought from the beginning
> that CREATE INDEX might be a showstopper for the whole HOT concept,
> and it's starting to look like that's the case.

I remember you raised this concern very early, but I am hopeful
that we would be able to solve this. Would it be acceptable
to have a simple (though not the best) solution for this release
and then improve later on ? As I mentioned earlier, one option
is to CHILL the table, if required, holding AccessExclusive lock,
just like VACUUM FULL. I am assuming here that CREATE INDEX is
not such a common activity, isn't that true ?

> I think what we need to get away from is the assumption that HOT-ness
> for one index is the same as HOT-ness for all. What if we only applied
> HOT to primary-key indexes, so that there was certainly not more than
> one index per table that the property applies to?
>

I think that will take away the ability to reuse HEAP_ONLY tuples
without vacuuming the heap and index.

Thanks,
Pavan

--

EnterpriseDB http://www.enterprisedb.com