Dave Page a écrit :
> The code is written, and is good as far as we are aware, but has not been through a beta/release cycle yet.
OK, that's already pretty good then.
Are there binary releases available ?
>> Is there a simple replication solution for windows then ? Or will
>> I have to stop the master postgresql at night to pgdump and
>> pgrestore on the backup server ?
>
> You don't need to stop the master server to pg_dump (in fact, it won't work if you do). On the slave, you can just
dropthe DBs and pg_restore them.
> PITR is another solution you might consider.
It looks quite harder to script though... But I'll consider it as
a better solution.
> Might be better if the users know they are being switched over - that way they will know to check for lost
transactionsetc. If you're using an ODBC app, consider using a separate DSN for the live and backup servers - similar
solutionswould apply to other interfaces of course.
There are no critical transactions here.
If the DB server fails and users are switched over to a backup
server with data of the day before, it is just fine.
Losing 1/2 of work is OK, but not working for 4 days is not.
So I might just write a script so that the person in charge when I
am away can run it and alert everyone.
--
Arnaud