On Mon, 2006-04-03 at 09:55 -0400, Tom Lane wrote:
> Simon Riggs <simon@2ndquadrant.com> writes:
> > Thinking about this some more, I ask myself: why is it we log index
> > inserts at all? We log heap inserts, which contain all the information
> > we need to replay all index inserts also, so why bother?
> I don't see any workable half measures.
Yep, looks that way.
> (1) We can't run user-defined functions during log replay. Quite
> aside from any risk of nondeterminism, the normal transaction
> infrastructure isn't functioning in that environment.
Didn't think of that one, but we could special case it.
> (2) Some of the index code is itself deliberately nondeterministic.
> I'm thinking in particular of the move-right-or-not choice in
> _bt_insertonpg() when there are many equal keys, but randomization is
> in general a useful algorithmic technique that we'd have to forswear.
Understood.
> (3) In the presence of concurrency, the sequence of heap-insert WAL
> records isn't enough info, because it doesn't tell you what order the
> index inserts occurred in. The btree code, at least, is sufficiently
> concurrent that even knowing the sequence of leaf-key insertions isn't
> full information --- it's not hard to imagine cases where decisions
> about where to split upper-level pages are dependent on which process
> manages to obtain lock on a page first.
In the presence of concurrency, it could be the OS who decides who gets
first crack at a page because of scheduling. We can never assume
anything there, so definitely a killer argument.
But that does open up an opportunity if we locked the table
AccessExclusive for a COPY LOCK into an existing table with indexes. One
day maybe, but not interesting enough yet.
Best Regards, Simon Riggs
