Re: including backend ID in relpath of temp rels - updated patch

On Thu, Aug 12, 2010 at 5:29 PM, Alvaro Herrera
<alvherre@commandprompt.com> wrote:
> Excerpts from Robert Haas's message of jue ago 12 13:29:57 -0400 2010:
>
>> We have two existing mechanisms for removing the catalog entries: when
>> a backend is first asked to access a temporary file, it does a DROP
>> SCHEMA ... CASCADE on any pre-existing temp schema.  And a table is in
>> wraparound trouble and the owning backend is no longer running,
>> autovacuum will drop it.  Improving on this seems difficult: if you
>> wanted to *guarantee* that the catalog entries were removed before we
>> started letting in connections, you'd need to fork a backend per
>> database and have each one iterate through all the temp schemas and
>> drop them.  Considering that the existing code seems to have been
>> pretty careful about how this stuff gets handled, I don't think it's
>> worth making the whole startup sequence slower for it.  What might be
>> worth considering is changing the autovacuum policy to eliminate the
>> wraparound check, and just have it drop temp table catalog entries for
>> any backend not currently running, period.
>
> What about having autovacuum silenty drop the catalog entry if it's a
> temp entry for which the underlying file does not exist?

I think that would be subject to race conditions.  The current
mechanism is actually pretty good, and I think we can build on it if
we want to do more, rather than inventing something new.  We just need
to be specific about what problem we're trying to solve.

--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise Postgres Company