On Thu, Jan 19, 2023 at 5:51 PM Andres Freund <andres@anarazel.de> wrote:
> I don't agree. But mainly my issue is that the devil you know (how this has
> worked for a while) is preferrable to introducing an unknown quantity (your
> patch that hasn't yet seen real world exposure).
Yeah, this is a major reason why I'm very leery about changes in this
area. A lot of autovacuum behavior is emergent, in the sense that it
wasn't directly intended by whoever wrote the code. It's just a
consequence of other decisions that probably seemed very reasonable at
the time they were made but turned out to have surprising and
unpleasant consequences.
In this particular case, I think that there is a large risk that
postponing auto-cancellation will make things significantly worse,
possibly drastically worse, for a certain class of users -
specifically, those whose vacuums often get auto-cancelled. I think
that it's actually pretty common for people to have workloads where
something pretty close to all of the autovacuums get auto-cancelled on
certain tables, and those people are always hard up against
autovacuum_freeze_max_age because they *have* to hit that in order to
get any vacuuming done on the affected tables. If the default
threshold for auto-cancellation goes up, those people will be
vacuuming even less often than they are now.
That's why I really liked your idea of decoupling auto-cancellation
from XID age. Such an approach can still avoid disabling
auto-cancellation just because autovacuum_freeze_max_age has been hit,
but it can also disable it much earlier when it detects that doing so
is necessary to make progress.
--
Robert Haas
EDB: http://www.enterprisedb.com