Christopher Browne wrote:

> In order to promote slave to master, there are two methods:
>
>1.  If the master is "dead", then you use FAIL OVER, discarding it. 
>Furthermore, if you only have two nodes, there's not much to preserve, 
>so I'd be inclined to simply clean out Slony-I stuff via UNINSTALL NODE.
>
>  
>
>2.  If the master is *not* to become dead, then you want to use MOVE SET
>to request that the origin be moved to the other node.  This requires
>that both nodes (indeed, it more or less requires  that *ALL* nodes in
>the cluster) be up.
>
>  
>

In the case when you plan to restart from scratch the master server, 
i.e. reinstall os, upgrade postgres, slony, etc (not a catastrofic 
crash, a planned os's upgrade and software versions instead), both 
option are  needed , arent they?
i mean, do a controlled move set and drop master node?
If so, what happen with the new before master now slave node, can i 
resubscribe it  with id 1 again?

---
Miguel