On Fri, Mar 6, 2009 at 9:47 PM, Vick Khera <vivek at khera.org> wrote:
> On Fri, Mar 6, 2009 at 1:42 AM, Stuart Bishop <stuart at stuartbishop.net> wrote:
>> The database is a freshly restored dump of the master node.
>
> I do not think you can do that.  You need to strip the slony stuff
> before you do the restore (or after).  Then re-enable replication by
> adding it as a node.

You can't strip the Slony stuff before the restore when doing disaster
recovery. The last time this came up on this mailing list, it seemed
this was the correct procedure (closely followed by 'just drop the
slony schema' which was the most popular mechanism, but apparently
doesn't quite do everything correctly).

We do a full dump every day, and test the restore by using the dump to
rebuild our staging server. So far, the process has worked fine. This
time... I don't know. To me it looks like I was just unlucky and hit a
case where repair config fails (which would be a bug).

-- 
Stuart Bishop <stuart at stuartbishop.net>
http://www.stuartbishop.net/