"William Meloney" <bmeloney@mindspring.com> writes:
> Developer dropped ~30+ tables from 4 different DBs on the same
> server. Queiries began to fail: unable to locate file;
> /var/lib/pgsql/pg_clog/0006.
>>
>> Hmm. What files do you actually have in /var/lib/pgsql/pg_clog?
> [ info snipped ]
Okay. It seems quite clear that you've executed less than a million
transactions so far in this installation; so 0000 is the only clog
segment that does or should exist, and there's no valid reason
for anything to be trying to access segment 0006. I do not think
this could be a clog logic error as I first suspected; it's probably
more like a garden-variety data corruption problem, ie, something
dropped a bit and there's now an xmin or xmax field someplace that
contains a bogus transaction ID up in the six-million-something range.
My advice is to try to track down and get rid of the bad tuple.
regards, tom lane
