As I understand it (and as stated in the README.upgrade file), all
slony versions
across a cluster have to be the same. Just to clarify it for me, this means that
upgrading from postgresql 7.4.2 / slony 1.0.5 to postgresql 8.1 / slony 1.1.5
involves:

- stop slony
- install the new version
- start slony
- prepare a new server (I also want a hardware upgrade), with new postgresql
  and the new slony version
- add the new server to the existing cluster
- once the new server is in sync, stop the application updating the database,
  and switch the origin to the new server.
- point the application to the new server (in my case, move the IP address
  used for this) and start it
- at this point the origin is running the versions I want
- now I can either remove one of the old nodes and reinstall it with the
  appropriate versions of postgresql / slony, or take a new server (for the
  hardware upgrade) and use that.
- Once the additional node is in sync, I can drop the old node(s) and I'm done.

Any comments? Doing things like upgrading slony on a DB server that's quite
critical always worries me :-)

Oh yeah, the versions of postgresql and slony I'm upgrading to is
pretty much determined by the versions available in Debian. I built
slony 1.0.5 myself back then, but I'd like to be able to use standard
package versions from now on.


Thanks,
Paul Slootman
_______________________________________________
Slony1-general mailing list
[email protected]
http://gborg.postgresql.org/mailman/listinfo/slony1-general

Reply via email to