Re: ALTER TABLE lock strength reduction patch is unsafe

On Mon, Jun 20, 2011 at 2:14 PM, Robert Haas <robertmhaas(at)gmail(dot)com> wrote:

> If this is a pre-existing bug, then it's not clear to me why we need
> to do anything about it at all right now.  I mean, it would be nice to
> have a fix, but it's hard to imagine that any proposed fix will be
> low-risk, and I don't remember user complaints about this.  I continue
> to think that the root of the problem here is that SnapshotNow is Evil
> (TM).  If we get rid of that, then this problem goes away, but that
> strikes me as a long-term project.

There are 2 bugs, one caused by my patch in 9.1, one that is pre-existing.

The 9.1 bug can be fixed easily. I will edit my patch down and repost
here shortly.

The pre-existing bug is slightly harder/contentious because we have to
lock the name of a possible relation, even before we know it exists.
I've been looking to implement that as a lock on the uint32 hash of
the relation's name and namespace. I'm looking for opinions ranging
from fix-now-and-backpatch thru to ignore and discuss for 9.2.

--
 Simon Riggs                   http://www.2ndQuadrant.com/
 PostgreSQL Development, 24x7 Support, Training & Services