Tom Lane wrote:
> I'll bet they're all trying to do anti-wraparound vacuums. This would
> explain why the problem hasn't been observed often enough to have been
> fixed long since.
Ah, that should have been obvious, shouldn't it. I'm able to reproduce
the problem by moving relfrozenxid backwards for a shared catalog, and I
can confirm that it works correctly with the patch proposed, so I have
pushed it to all branches. (I only tested the problem scenario in 9.1
and master, since it's too much manual work).
Thanks for the report!
--
Álvaro Herrera http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services