Re: odd problem !

Oleg Bartunov <oleg(at)sai(dot)msu(dot)su> writes:
> below is the problem I just bitten when play with toy db. I did:

> 1.initdb -D ./t1
> 2. pg_ctl -D ./t1 start
> 3. createdb test
> 4. psql test -c "create table a (f integer);"
> 5. run script which populates table a in background
> perl bgupdate.pl &
> 6. cp -a ./t1 ./t2

I would not really expect this to produce a usable copy at all...

> 6. pg_ctl -D ./t1 stop
> it's waiting for bgupdate.pl, so I killed script
> 7. pg_ct -D ./t2 start
> see, that everything is fine.

Exactly how much did you test? However, that doesn't seem relevant
to your subsequent problem with the original.

> 8. pg_ctl -D ./t2 stop
> 9. pg_ctl -D ./t1 start
> 10.

> pg(at)zen:~/test$ psql -l
> List of databases
> Name | Owner | Encoding
> -----------+----------+----------
> template0 | postgres | KOI8
> template1 | postgres | KOI8
> test | postgres | KOI8
> (3 rows)

> 11.
> pg(at)zen:~/test$ psql test
> FATAL: database "test" does not exist
> psql: FATAL: database "test" does not exist

What PG version is this exactly? I suppose that you're seeing one of
the GetRawDatabase corner cases, but I'm not quite sure why the shutdown
and restart after the createdb wouldn't have fixed it. Does vacuuming
pg_database and then checkpointing make it work?

regards, tom lane