On Wed, May 16, 2018 at 09:33:46AM -0400, Robert Haas wrote:
> > In sum, I think the problem is mostly solved. Backend 2 unpins the
> > segment in next ts_lexize() call. But if backend 2 doesn't call
> > ts_lexize() (or other TS function) anymore the segment will remain mapped.
> > It is the only problem I see for now.
>
> Maybe you could use CacheRegisterSyscacheCallback to get a callback
> when the backend notices that a DROP has occurred.
Yes, it was the first approach. DSM segments was unpinned in
InvalidateTSCacheCallBack() in that approach, which is registered using
CacheRegisterSyscacheCallback().
I haven't deep knowledge about guts of invalidation callbacks. It seems
that there is problem with it. Tom pointed above:
> > I'm not sure that I understood the second case correclty. Can cache
> > invalidation help in this case? I don't have confident knowledge of cache
> > invalidation. It seems to me that InvalidateTSCacheCallBack() should
> > release segment after commit.
>
> "Release after commit" sounds like a pretty dangerous design to me,
> because a release necessarily implies some kernel calls, which could
> fail. We can't afford to inject steps that might fail into post-commit
> cleanup (because it's too late to recover by failing the transaction).
> It'd be better to do cleanup while searching for a dictionary to use.
But it is possible that I misunderstood his note.
--
Arthur Zakirov
Postgres Professional: http://www.postgrespro.com
Russian Postgres Company
