On Sat, Sep 17, 2005 at 07:41:18PM +0800, Ivan R. Sy Jr. wrote:
> hi all
> 
> i was able to simulate 5 different boxes using FreeBSD 5.4 and slony1.1 
> Perltools installed using  PerlTools for Slony 1.1 guide. 
> http://www.varlena.com/GeneralBits/113
> 
> Ive tested controlled swithover and it worked fine. I've also tried 
> hostile failover as well and it's neat.
> 
> But i have a problem on rejoining the original master back to the 
> cluster. For example:
> 
> BoxA is master and BoxB,C and D are slaves
> BoxA suddenly dies, then BoxB takes over no problem
> (i got the script for that)
> but when BoxA is available, how can it rejoin and become the master 
> again? perhaps someone there know a sample script that joins the processes?
> 
> someone adviced me to do this:
> 
> -drop the dead node on BoxA
> (failover script in BoxB)
> 
> -recreate the empty database on BoxA
> (i need guide, sample script? uninstall? init cluster?)

dropdb BoxA; createdb BoxA; 
load the empty schema (without slony) (you have that right?) 
onto BoxA.  (Same way you added other replicas.)

> 
> -Add the node to the cluster
> (i need guide, sample script? no idea how)

Perltools has an add_node script

> 
> -Subscribe the node to the set
> (i have a script for this, but any suggestion is welcome)
> 
Perltools has an subscribe_node script


When you failed over, did you verify that your downstream
databases were all replicating from the new origin?

Did you name your nodes so that you are not using node 1?

--elein


> 
> 
> Thanks All!
> 
> 
> 
> _______________________________________________
> Slony1-general mailing list
> Slony1-general at gborg.postgresql.org
> http://gborg.postgresql.org/mailman/listinfo/slony1-general
>