Heikki Linnakangas <[EMAIL PROTECTED]> writes:
> On Sun, 8 May 2005, Tom Lane wrote:
>> While your original patch is buggy, it's at least fixable and has
>> localized, limited impact. I don't think these schemes are safe
>> at all --- they put a great deal more weight on the semantics of
>> the filesystem than I care to do.
> I'm going to try this some more, because I feel that a scheme like this
> that doesn't rely on scanning pg_class and the file system would in fact
> be safer.
I think this proposal is getting more and more invasive and expensive,
and it's all to solve a problem that we do not even know is worth
spending any time on. I *really* think this is the wrong direction
to take. Aside from the required effort and risk of breaking things,
the original patch incurred cost only during crash recovery; this is
pushing costs into the normal code paths.
> Delay the actual file creation until it's first written to. The write
> needs to be WAL logged anyway, so we would just piggyback on that.
This is a bad idea since by then it's (potentially) too late to roll
back the creating transaction if the creation fails. Consider for
instance a tablespace directory that's mispermissioned read-only, or
some such. I'd rather have the CREATE TABLE fail than a later INSERT.
(Admittedly, we can't completely guarantee that an INSERT won't hit
some kind of filesystem-level problem, but it's still something to
try to avoid.)
Also, the "first write" actually comes from mdextend, which is not a
WAL-logged operation AFAIR. Some rethinking of that would be necessary
before this would have any chance of working.
regards, tom lane
---------------------------(end of broadcast)---------------------------
TIP 6: Have you searched our list archives?
http://archives.postgresql.org
