Re: advance local xmin more aggressively

Alvaro Herrera <alvherre(at)commandprompt(dot)com> writes:
> For example, maybe we could keep track of counts of snapshots removed
> since the last xmin calculation, and only run this routine if the number
> is different from zero (or some small positive integer).

I think most of the callers of SnapshotResetXmin already know they
removed something.

It might be interesting for FreeSnapshot or something nearby to note
whether the snapshot being killed has xmin = proc's xmin, and only do
the update calculation if so.

I still dislike the assumption that all resource owners are children of
a known owner. I suspect in fact that it's demonstrably wrong right
now, let alone in future (cf comments in PortalRun). If we're going to
do this then snapmgr.c needs to track the snapshots for itself. Of
course that's going to make the "is it worth it" question even more
pressing.

regards, tom lane