On Wed, Apr 9, 2008 at 12:43 PM, Christopher Browne <
cbbrowne at ca.afilias.info> wrote:

> "Shoaib Mir" <shoaibmir at gmail.com> writes:
> > Does anyone has any idea about it? I am looking for any Slony
> > catalog table differences that I couldnt find myself in this
> > specific scenario.
>
> It depends heavily on why the node failed.
>
> - If it failed because the DBMS got corrupted, then there could be
>  pretty well any sort of trash in the DB on the failed node.
>
> - If you chose to fail over because [say] there was a temporary
>  communications problem, then that database essentially gets "shunned."
>
>  It will appear, internally, to be in perfectly good shape, but, as
>  you noticed, the cluster wasn't quite happy with it ;-(.
>
> You really need to drop the node and recreate it.
>
> It's not worth trying to reconstruct that node into the cluster; head
> back to the problem statement, namely:
>
>  "We had such a serious problem that it warranted doing a FAIL OVER."
>
> If you had so serious a problem, then it warrants rebuilding that
> database from scratch, period.
>

Thanks a lot for a detailed reply on this, this was helpful. I was also
finally able to find the differences in sl_subscribe tables for failed node
and active nodes which will help me to identify the failed nodes that have
not yet been deleted from cluster.

Regards,
Shoaib
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.slony.info/pipermail/slony1-general/attachments/20080409/=
d8373d74/attachment-0001.htm