On Apr 30, 2013, at 11:14 PM, Yang Zhang wrote:
> Is this the extent of what I can expect, *always*, even if I had run
> the proper experiment involving pulling the cord (or at least kill
> -9)?
I would not count on it. And if it works 100% reliably now, it might not on a future version of PostgreSQL.
As Josh Berkus pointed out to my off-list, there are two competing definitions of the term "recover" in use here:
1. In my blog post, the definition of "recover" was "bring up the database without having unusually extensive knowledge
ofPostgreSQL's internals."
2. For Tom, the definition of "recover" is "bring up the database if you have appropriate knowledge of PostgreSQL's
internals."
You can't recover from the lost of a tablespace per definition #1. You can per definition #2.
I'd strongly suggest that relying on definition #2, while absolutely correct, is a poor operational decision for most
users.
--
-- Christophe Pettus
xof@thebuild.com