> I think
> we need a detailed design document for how this is all going to work.
> We need to not only handle the master properly but also handle the
> slave properly. Consider, for example, the case where the slave
> begins to replay the transaction, reaches a restartpoint after
> replaying some of the new pages, and then crashes. If the subsequent
> restart from the restartpoint blows away the main relation fork, we're
> hosed. I fear we're plunging into implementation details without
> having a good overall design in mind first.
As I said in my first post, I'm basing the patch on the post:
http://archives.postgresql.org/pgsql-hackers/2011-01/msg00315.php
So I assumed the design was ok (except for the "stray file around
on a standby" case, which has been discussed earlier on this thread).
If there are things to be discussed/analyzed (I guess the restart point
thing is one of those) we can do it... but I thought that the whole
design was somehow in place
Leonardo