Has anyone ever seen a database spontaneously change character sets? I
could have sworn that all databases in my replication slave database
cluster were set up as SQL_ASCII. Now they are all UNICODE, and slony
is failing to replicate, due to invalid byte sequences.
Now I can't really *swear* that the databases weren't UNICODE all along
but I do know for sure:
- that I intended them to be SQL_ASCII
- that I would make an effort to avoid UNICODE until we upgrade to 8.1
- that slony has been replicating without problem on this database for
some months
The only unusual thing that has been done with that database recently is
a failed attempt to upgrade slony from 1.1 to 1.5. This was about 3
days before slony ran in to character set problems.
I have looked through the logs and found nothing of interest.
Any ideas?
This is non-production, on PostgreSQL 8.0.3
__
Marc