Re: sync_file_range()

"Tom Lane" <tgl(at)sss(dot)pgh(dot)pa(dot)us> writes:
> > Indeed, I've been wondering lately if we shouldn't resurrect
> > LET_OS_MANAGE_FILESIZE and make that the default on systems with
> > largefile support. If nothing else it would cut down on open/close
> > overhead on very large relations.
>
> > I'd still put some limit on the filesize, else you cannot manually
> > distribute a table across spindles anymore. Also some
> backup solutions
> > are not too happy with too large files eighter (they have
> trouble with
> > staging the backup). I would suggest something like 32 Gb.
>
> Well, some people would find those arguments compelling and
> some wouldn't. We already have a manually configurable
> RELSEG_SIZE, so people who want a 32Gb or whatever segment
> size can have it.
> But if you're dealing with terabyte-sized tables that's still
> a lot of segments.
>
> What I'd be inclined to do is allow people to set RELSEG_SIZE
> = 0 in pg_config_manual.h to select the unsegmented option.
> That way we already have the infrastructure in pg_control etc
> to ensure that the database layout matches the backend.

That sounds perfect. Still leaves the question of what to default to ?

Another issue is, that we would probably need to detect large file
support of the underlying filesystem, else we might fail at runtime :-(

Andreas