Re: Unexpected VACUUM FULL failure

"Tom Lane" <tgl(at)sss(dot)pgh(dot)pa(dot)us> writes:

> The reason why XLogAsyncCommitFlush() is ugly is that it doesn't
> actually accomplish a darn thing, as we now see from this failure:
> it does not guarantee that hint bits will get set, because of the
> inexact bookkeeping in clog.c. It might marginally improve the
> probability that they get set, but that's all. The reason I want
> to take it out is mainly that its very existence tempts people to make
> the same mistake that was made here.

I don't understand your reasoning here and I would like to understand it so if
you don't mind I would like to see if I can work out what you're talking
about. Regardless of this point I think the impact of what you were proposing
to do to VF instead was much less than Simon seemed to think it was so it
seems like a perfectly acceptable solution.

As far as I understand the Xlog flush in combination with keeping an exclusive
lock on table and always holding locks until the end of the transaction make
forcing the hint bits entirely safe.

The fragility you see comes from depending on how those three things interact
and the difficulty in ensuring that all of those properties are always true.
By "marginally improve the probability" you're making a judgement of the
probability that programmers will manage to maintain all those properties
consistently, not about the probability that the race will occur at run-time?

--
Gregory Stark
EnterpriseDB http://www.enterprisedb.com