Just keeping it in a separate email, incase this is thrashed down.
vacuum full has a lot of problem stories, not just because the db gets locked, but also because it is mostly (mis)used when there are space issues.
of course, there are strong warnings in docs and wiki about using a vacuum full,
but i was just thinking
vacuumdb and vacuumlo can make use of options like --dry-run, like pg_repack allows,
but also just abort incase there is no disk space available.
i mean, i have seen confident replies from earlier mailing threads, where assume 2x disk required per object that is full vacuumed at least.
can this be converted into an estimate of sorts on objects that would be vacuumed and free disk that would be required before any operation starts.
maybe this is wishful thinking, but just asking.