Michael Glenn <mike(at)mglenn(dot)com> writes:
> The log is only 98K.
Okay, then you haven't executed four billion transactions...
Probably the next thing is to look at pg_database and try to understand
why all the rows are showing as dead. There are a couple of tools
around that can dump out the contents of tuple headers. I'd suggest
pg_filedump (http://sources.redhat.com/rhdb/tools.html) but it is
intended for 7.1/7.2 and you'd need to do a little surgery to make it
work on 7.0 files. (I think ripping out the code that expects an LSN
field to be present in the page header would be enough.) If you check
the PG list archives you will find references to one or two older
programs in the same vein; there might still be something available
that will work on 7.0 files without modification.
regards, tom lane