Hello all,
We have a live server running on PG 8.0.3 which we backup on a regular bases
by doing a pg_dumpall --oids. We have a test server currently running 8.1.1
where we reload that dumpall using psql.
Now we have a history tracking system in the db which via triggers and rules
keeps track of changes to certain tables via OID and logs them in history
tables. Now our problem is that when we load the db from 8.0.3 into the
8.1.1 db, the current data with all the oids is loaded correctly. But, as
soon as we try to add rows (new data created during testing), we run into a
problem where the OID on the testing machine has collisions with data that
was loaded from the live server.
I've been searching for awhile on further information about OID and haven't
been real successfull. Any pointers would be helpfull...
1. Is OID stored per cluster, or is it per database? I assume it's per
cluster since databases appear to have an OID.
2. Is there any way to get the test machine to have it's OID reset to the
current live server OID after a reload so we don't get OID collisions?
We aren't sure at this point if the problem we are having has to do with the
fact that we drop and reload the test db quite often and so we are running
into OID wraparound, or if it has to do with the fact that we recently went
to 8.1 on this test machine and OID is now handled differently on reload.
Anyt thoughts or suggestions?
-Chris
