Re: proposed TODO: non-locking CREATE INDEX / REINDEX

On R, 2005-06-10 at 17:54 +0300, Hannu Krosing wrote:
> On R, 2005-06-10 at 09:47 -0400, Tom Lane wrote:
> > Hannu Krosing <hannu(at)skype(dot)net> writes:
> > > There are many ways this could be made to work, so it needs some
> > > discussion.
> >
> > (1) when do you ever catch up?
> >
> > (2) if your answer to (1) involves increasing the strength of a lock,
> > how do you avoid risk of deadlock?
>
> No. I don't plan on locking the table at all.
>
> The only thing that is changed during the initial fast-build-index is
> that new tuples are inserted after CTID_INDEX_MIN, and after the initial
> fastbuild index is done, the only restriction is that the index can't be
> used in queries before the tuples between CTID_INDEX_MIN and
> CTID_INDEX_MAX are added to the index.

Maybe I did not make it very clear, but the initial fast-build-index is
done only for the tuples whose ctid is below CTID_INDEX_MIN, thereby
this also will happen in finite amount of time.

The total time will be:
initial_index_build_time +
inserting_tuples_added_while_doing_the_initial_build

the latter wil be slowed down a little, as these are competing with
index entries from inserts happening in real time.

> As the number of tuples between CTID_INDEX_MIN and CTID_INDEX_MAX is
> finite, they must be added in finite time, by which time the index will
> be up-to-date and usable for querie planner. (i.e. (1) is done)
>
> All tuples inserted after the initial fast-build-index has finished and
> CTID_INDEX_MAX is fixed, are inserted into the index at the time of
> inserting the tuple, like for any other index.
>
--
Hannu Krosing <hannu(at)tm(dot)ee>