Re: Are folks running 8.4 in production environments? and 8.4 and slon 1.2?

Tory M Blue wrote:
>
> Torsten Zühlsdorff wrote:
>
> >
> > I've upgraded all my databases to 8.4. They pain was not so big, the
> > new -j Parameter from pg_restore is fantastic. I really like the new
> > functions around Pl/PGSQL. All is stable and fast.
> >
> > Greetings from Germany,
> > Torsten
>
>
> On Sat, Oct 17, 2009 at 8:55 PM, Karl Denninger <karl(at)denninger(dot)net
> <mailto:karl(at)denninger(dot)net>> wrote:
>
> I am running all my production work on 8.4 at this point; no
> problems of
> note.
>
> -- Karl
>
>
> Thanks guys, ya running slon , must be replicated so it seems 8.4 is
> good, but need to wait for the slon+pg to be good..
>
> Also not real pain? A full dump and restore again, can't see that not
> being painful for a DB of any real size.
>
> Thanks for the response guys.
>
> Und Danke Torsten :)
>
> Tory
I am running Slony on 8.4; it complains on init but I have checked it
EXTENSIVELY and it is replicating fine. I have a distributed forum
application that uses Slony as part of the backend architecture and it
would choke INSTANTLY if there were problems.

Didn't dump and restore. This is how I migrated it:

1. Set up 8.4 on the same machine.
2. Use slony to add the 8.4 "instance" as a slave.
3. Wait for sync.
4. Stop client software.
5. Change master to the 8.4 instance.
6. Shut down 8.3 instance, change 8.4 instance to the original port
(that 8.3 was running on) and restart.
7. Bring client application back up.
8. Verify all is ok, drop the 8.3 instance from replication.

Total downtime was about 2 minutes for steps 4-8, the long wait was for
#3, which has no consequence for the clients.

Note that this requires 2x disk storage + some + enough I/O and CPU
bandwidth to get away with the additional replication on the master
machine. If you don't have that you need a second machine you can do
this to and then swap the client code to run against when the
replication is complete.

-- Karl