On Wed, Feb 16, 2022 at 1:04 PM Robert Haas <robertmhaas@gmail.com> wrote:
> No, what I'm saying is that people running older versions routinely
> run VACUUM in single-user mode because otherwise it fails due to the
> truncation issue. But once they go into single-user mode they lose
> protection.
Seems logically consistent, but absurd. A Catch-22 situation if ever
there was one.
There might well be an element of survivorship bias here. Most VACUUM
operations won't ever attempt truncation (speaking very generally).
How many times might (say) the customer that John mentioned have
accidentally gone over xidStopLimit for just a little while, before
the situation corrected itself without anybody noticing? A lot of
applications are very read-heavy, or aren't very well monitored.
Eventually (maybe after several years of this), some laggard
anti-wraparound vacuum needs to truncate the relation, due to random
happenstance. Once that happens, the situation is bound to come to a
head. The user is bound to finally notice that the system has gone
over xidStopLimit, because there is no longer any way for the problem
to go away on its own.
--
Peter Geoghegan