Re: Question about ALTER TABLE SET TABLESPACE locing

Ühel kenal päeval, K, 2006-02-01 kell 18:08, kirjutas Tom Lane:
> Hannu Krosing <hannu(at)skype(dot)net> writes:
> > Does ALTER TABLE SET TABLESPACE lock the table
>
> It had better ... see nearby discussion about relaxing locking for
> TRUNCATE.

Is it some recent disussion ?

> Exactly the same problem that at commit we'd be cutting
> the ground out from under any concurrent query.

Can't we just keep READ locks and reopen the datafile from its new
location before releasing it ?

Does our shared memory page cache implementation track logical or
physical pages ? If it's just logical pages, then moving the physical
storage around below should not affect it.

Of course there are problems with WAL/PITR which *have* to deal with
physical storage, but this should be mostly unrelated.

--------------
Hannu