Alvaro Herrera <alvherre@commandprompt.com> writes:
> In fact I don't understand what's the point about multiple databases vs.
> a single database. Surely a checkpoint would flush all buffers in all
> databases, no?
Yeah --- all the ones that are dirty *now*. Consider the case where you
vacuum DB X, update its datvacuumxid, and don't checkpoint because the
global min didn't advance. Now you crash, possibly leaving some hint
bits unwritten; but the datvacuumxid change did make it to disk. After
restart, vacuum DB Y, update its datvacuumxid, and find that the global
min *did* advance. You checkpoint, and that guarantees that DB Y is
clean for the clog truncation. But DB X isn't.
The 8.2 changes have created the equivalent risk at the level of each
individual table. We can't write a vacuumxid change unless we are sure
that the hint-bit changes it promises are actually down to disk.
regards, tom lane
