On 3/4/2005 3:35 PM, Alan Hodgson wrote:

> Is there a way to backup and restore a cluster member without having
> to redo the replication setup?  I have 5 nodes to upgrade to 8.0.1.
> The slave nodes only replicate a subset of the tables on the master.  
> Re-replicating out the slave nodes and setting up the paths and listeners,
> especially one at a time, is a major pain.
> 
> I know enough to kill the slons, make the db unavailable to other processes,
> and back it up and restore it without anything else touching it.  I know
> to update the table OID's in sl_table.  However, I seem to be missing where
> it sticks the old trigger info.  Is this even possible?
> 

<BLINK> WARNING WARNING WARNING </BLINK>

I have not tested any of the below! It is an idea that you have to prove 
right or wrong!

END WARNING

After "disconnecting" a subscriber, you want to call the 
alter_table_restore procedure for all replicated tables. That will 
remove the slony trigger and restore the original application triggers 
and rewrite rules.

Then you do the upgrade via dump+restore

Next you call the alter_table_for_replication procedure for all those 
tables to get back into subscriber status.

Last you invoke a slonik script that reloads the db version specific 
parts of the slony stored procedures.

The subscriber should now be ready to start its slon and catch up.


Good luck and let us know if it works that way.


Jan

-- 
#======================================================================#
# It's easier to get forgiveness for being wrong than for being right. #
# Let's break this rule - forgive me.                                  #
#================================================== JanWieck at Yahoo.com #