* Tom Lane <tgl@sss.pgh.pa.us> [010429 23:12] wrote:
> Bruce Momjian <pgman@candle.pha.pa.us> writes:
> > big problem is that there is no good way to make the symlinks reliable
> > because in a crash, the symlink could point to a table creation that got
> > rolled back or the renaming of a table that got rolled back.
>
> Yes. Have you already forgotten the very long discussion we had about
> this some months back? There is no way to provide a reliable symlink
> mapping without re-introducing all the same problems that we went to
> numeric filenames to avoid. Now if you want an *UNRELIABLE* symlink
> mapping, maybe we could talk about it ... but IMHO such a feature would
> be worse than useless. Murphy's law says that the symlinks would be
> right often enough to mislead dbadmins into trusting them, and wrong
> exactly when it would do the most damage to trust them. The same goes
> for other methods of unreliably exporting the name-to-number mapping,
> such as dumping it into a flat file.
>
> We do need to document how to get the mapping (ie, select relfilenode,
> relname from pg_class). But I really doubt that an automated method
> for exporting the mapping would be worth the cycles it would cost,
> even if it could be made reliable which it can't.
Perhaps an external tool to rebuild the symlink state that could be
run on an offline database. But I'm sure you have more important
things to do. :)
--
-Alfred Perlstein - [alfred@freebsd.org]
Daemon News Magazine in your snail-mail! http://magazine.daemonnews.org/
