On Thu, 2010-08-19 at 17:51 -0400, Christopher Browne wrote:
> Selena Deckelmann <selenamarie at gmail.com> writes:
> > It would be lovely if there was a way to get a list of potential locks
> > required by a script run, without actually taking the locks.
> 
[snip]
> Fortunately, you may an excellent approximation (which should be good
> enough for any but rather perverse situations) by doing a QA run (not in
> production!) as follows:
[snip]
> If there are crucial differences between the data QA and production,
> then the list *could* be wrong, due to the DDL behaving differently.
> That's an argument for realistic test environments, not for not doing
> this.

What about running the locking test on an actual replica (and perhaps
inside a transaction which is rolled back at the end) ? For any sane DDL
which should replicate correctly that should take the same locks as on
the master, and if you stop the slon daemon for that replica (or execute
the test _via_ the slon daemon) it will also be sure not to deadlock.

Cheers,
Csaba.