This feedback concerns Slony 2.1.2, being used with PostgreSQL 9.1.

We're using it on Amazon EC2, which ads the complication that the
internal IP address of the machine changes every time it starts up.

On Friday, our underlying hardware failed on the master, and we had to
stop/start the master DB instance, to move it to new hardware.

When slony started, the name of the server itself was still resolving to
the old internal IP, so this happened:

###
2012-11-09 21:35:20 EST FATAL  main: Cannot connect to local database -
could not connect to server: Con
nection timed out
        Is the server running on host "db.example.com" (10.79.78.214)
and accepting
        TCP/IP connections on port 5432?
###

After that, Slony was never able to correct itself, even long after the
TTL for the related DNS record would have expired. It took a manual
stop/start to get things going in.

Perhaps Slony aggressively caches DNS? Or perhaps it could recover from
this "FATAL" condition? Is this expected behavior, or should slony
behave better in the face of temporary DNS problems?

Thanks for insights!

   Mark