oids rollover?

Daniel Kalchev <daniel@digsys.bg> writes:

I have a problem with an 7.1.3 database that has probably overflowed
the oid counter. The startup halts with these messages

DEBUG: database system was interrupted at 2002-06-24 21:19:43 EEST
DEBUG: CheckPoint record at (156, 1692817164)
DEBUG: Redo record at (156, 1692775580); Undo record at (0, 0);
Shutdown FALSE
DEBUG: NextTransactionId: 859255800; NextOid: 7098
FATAL 2: Invalid NextTransactionId/NextOid
postmaster: Startup proc 4752 exited with status 512 - abort

Looks that way. This is fixed in 7.2, so you might want to think about
an update sometime soon.

Can something be sone to recover the database?

You could modify contrib/pg_resetxlog to force a value at least 16384 into
the OID counter. Since the DB was evidently not shut down cleanly, I'd
counsel cutting out the xlog-reset function entirely; just make it read
the pg_control file, set a valid nextOid, update the CRC, and rewrite
pg_control.

regards, tom lane