Re: 8.1 index corruption woes

Alvaro Herrera <alvherre@commandprompt.com> writes:
> I don't want to discard this idea, because we're getting a very
> unusually high number of bogus entries.  However, they are all (or a
> very high percentage of them) the very first entry on each index page.
> I want to confirm that the leftmost on a leaf btree page is a valid
> item, and not something like the lower bound value?  (I think we only
> store high bounds on internal pages, not leaf pages, but I'm not sure).

Er ... no.  Per nbtree/README:

: On a page that is not rightmost in its tree level, the "high key" is
: kept in the page's first item, and real data items start at item 2.
: The link portion of the "high key" item goes unused.  A page that is
: rightmost has no "high key", so data items start with the first item.
: Putting the high key at the left, rather than the right, may seem odd,
: but it avoids moving the high key as we add data items.

(Right offhand, it looks like _bt_split just copies the item that it's
cloning the index key of.  Maybe it would be worth setting the item
pointer invalid, to prevent confusion in future?)
        regards, tom lane