Re: preserving forensic information when we freeze

On 2013-12-20 07:12:01 -0500, Robert Haas wrote:
> I think the root of the problem is that nobody's very eager to add
> more hidden system catalog columns because each one bloats
> pg_attribute significantly.

I think that part is actually solveable - there's no real need for them
to be real columns, present in pg_attribute, things could easily enough be
setup during parse analysis. The bigger problem I see is that adding
more useful columns will cause name clashes, which will probably
prohibit improvements in that direction.

> Maybe what we should do is add a function something like
> pg_tuple_header(tableoid, ctid) that returns a record, maybe something
> like (rawxmin xid, rawxmax xid, rawcid cid, infomask int, infomask2
> int, hoff int). Or perhaps some slightly more cooked version of that
> information. And then delete the xmin, xmax, cmin, and cmax system
> columns. That'd save significantly on pg_attribute entries while, at
> the same time, actually providing more information than we do today.

I was wondering whether we couldn't just pass pg_tuple_header() a whole
row, instead of having the user manually pass in reloid and ctid. I
think that should actually work in the interesting scenarios.

> pageinspect is useful, too, but it seems to deal mostly with pages, so
> I'm not sure it's a natural substitute for something like this.

Agreed. I also think that we really need something to investigate
problems in core, not in a contrib module people won't have installed,
especially in bigger setups. That said I plan to either submit some
improvements to pageinspect myself, or convince somebody else to do so
in the not too far away future.

Greetings,

Andres Freund

--
Andres Freund http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services