Hans-Juergen Schoenig <postgres@cybertec.at> writes:
> What I am doing is DELETE -> COPY into serveral tables. The problem is:
> It seemed to work pretty fine up to now but now PostgreSQL stops in the
> middle of the COPY command. The CPU load goes down to zero but the
> process stays alive.
Bizarre.
> When I press CTRL + C the command line isn't
> blocked anymore but the COPY command starts to work as a background
> process again (100% CPU load).
Perhaps the backend's gotten into some kind of loop, and isn't reaching
the point at which it would realize that the client connection is gone
and then exit. Can you attach to the looping backend with a debugger
and get a stack trace for us?
> Also, I had some
> NOTICE: Rel pg_artikel: Uninitialized page 74 - fixing
> NOTICE: Rel pg_artikel: Uninitialized page 75 - fixing
> before ...
These are suspicious, but not hard evidence of a problem. IIRC we've
determined that there are legitimate cases where zero pages could get
inserted into a table. See past threads about "Uninitialized page"
messages.
regards, tom lane
