Re: orphaned PGDATA/base/ subdirectories

Reece Hart <reece(at)harts(dot)net> writes:
> The manual clearly says that a each row in pg_database will have a
> corresponding subdirectory in base/. Is the converse also true, ie,
> should every subdir in base/ also have a row in pg_database? If so, am
> I safe to rm the unreferenced subdirectories?

Yes.

> FWIW, I know that I've interrupted (^C) a few createdb -T operations
> during the course of this installation, and I recently had the disk full
> error. My best guess is that these orphaned directories result from
> those interruptions and failure.

Hm ... 8.1 and up contain code that tries to remove the copied files if a
CREATE DATABASE fails midway through, but it seems it's not working for
you.

It strikes me that if a control-C arrives while the thing is already
doing remove_dbtablespaces(), that would kick it out of that code,
allowing some files to go unreclaimed --- are you in the habit of
banging on ^C repeatedly? I couldn't reproduce such a problem in a
small amount of testing though.

regards, tom lane