On Fri, 2007-01-19 at 08:15 -0600, Jeff Amiel wrote:
> Yesterday the RAID failed in one of my subscriber nodes for my 
> production database.  I am in the process of restoring a pg_dumpall to 
> the box but am uncertain what to do about Slony.
> 
> The backup was made at 3:00am or so...but the raid failure didn't occur 
> until about 8:30am.  I am assuming that once my restore is complete and 
> I fire slony back up, that it won't have the trigger events to 
> re-populate the missing items in the database (how could it?).  
> Certainly its not possible that slony will  analyze the tables for 
> differences and generate new events, correct? 
> 
> How should I get this subscriber back in sync?  Should I drop nuke all 
> the data in the subscriber, drop the node and recreate and let it start 
> from scratch?

You don't have much of a choice here.  When you add a node to
replication, slony deletes (or truncates) all the data in the target
tables and repopulates them from scratch.

What you want to do is:

-Forget about restoring the data on the subscriber (unless it contains
stuff that isn't replicated that you need).
-extract the schema only (no data) from the master database.
-restore that on the subscriber
-add the node back to replication and subscribe it.

-- 
Brad Nicholson  416-673-4106
Database Administrator, Afilias Canada Corp.