On Thu, 21 Jun 2001, Nathan Myers wrote:
> > I suppose that it rather depends on how you expected to
> > move the logs over. My approach was to archive the redo
> > when PG is done with them and only then to roll them
> > forward.
> The above makes no sense to me. A hot recovery that discards some
> random number of committed transactions is a poor sort of recovery.
Agreed. Nevertheless it's at least db_size - 1Mb better
than the current options.
Recovering the rest manually from log files is good enough
for us (indeed, much better than the potential loss of
performance or reliability from "real" replication).
If it horrifies you that much, think of it as 15-minutely
incremental backups.
Matthew.