On 11/17/2010 04:25 PM, Tom Lane wrote:
> I'm afraid that any such change would trade a visible, safe failure
> mechanism (no avworker) for invisible, impossible-to-debug data
> corruption scenarios (due to failure to reset some bit of cached state).
> It certainly won't give me any warm fuzzy feeling that I can trust
> autovacuum.
Well, Alvaro doesn't quite seem have a warm fuzzy feeling with the
status quo, either. And I can certainly understand his concerns.
But yes, the os-level process separation and cache state reset guarantee
that an exit() / fork() pair provides is hard to match up against in
user space.
So, Alvaro's argument for robustness only stands under the assumption
that we can achieve a perfect cache state reset mechanism. Now, how
feasible is that? Are there any kind of tools that could help us check?
Regards
Markus Wanner