Re: Re: [COMMITTERS] pgsql: Augment WAL records for btree delete with GetOldestXmin() to

On Sun, 2010-01-31 at 16:53 -0500, Tom Lane wrote:
> Heikki Linnakangas <heikki(dot)linnakangas(at)enterprisedb(dot)com> writes:
> > IIRC it was Greg Stark who suggested last time this was discussed that
> > we could calculate the exact value for latestRemovedXid in the standby.
> > When replaying the deletion record, the standby could look at all the
> > heap tuples whose index pointers are being removed, to see which one was
> > newest.
>
> This assumes that the standby can tell which heap tuples are being
> removed, which I'm not sure is true --- consider the case where the
> deletion record is carrying a page image instead of a list of deleted
> tuples. But maybe we could rejigger the representation to make sure
> that the info is always available. In general it sounds like a much
> nicer answer.

I'm looking at this now since we definitely need either this or the
previous patch.

XLOG_BTREE_DELETE records would not carry latestRemovedXid, this would
be calculated by inspection of heap tuples. XLOG_BTREE_DELETE would have
a heap relid, to allow direct access to the heap without a Relation. We
would check xmin, xmax and xvac. Code will look a little like
heap_fetch() though disimilar enough not to use directly. We would only
look at the tuple header and info.

--
Simon Riggs www.2ndQuadrant.com