Re: I'm puzzled by a foreign key constraint problem

Jonathan Guthrie <jguthrie@brokersys.com> writes:
> ... or, at least, it's supposed to.  Those two operations are not supposed
> to overlap at all even if they're on two different connections.  I
> thought I had verified this by looking at the log file.  I mean, I can
> look at the log file and see things like

> 2008-11-03 16:29:22 CST DEBUG:  00000: StartTransactionCommand
> and
> 2008-11-03 16:29:22 CST DEBUG:  00000: CommitTransactionCommand

> where I would expect them to if what I'm expecting is going on, but the
> log file doesn't appear to have enough information to see a transaction
> created, proceed, and then end.  That is, how do I know which
> transaction was started and which one was committed?

You need to add more identification info to your log_line_prefix.
The PID would be the most reliable way to tie those entries together,
but I think there's also an option that writes the transaction ID.

> I'm kind of confused by lines like this:

> 2008-11-03 16:29:22 CST DEBUG:  00000: name: unnamed; blockState:    INPROGRESS; state: INPROGR, xid/subid/cid:
678145/1/4,nestlvl: 1, children: 678146 678147 

> Is there an easy explanation somewhere?

You'd have to look at the source code to figure out most of the
DEBUG-level messages.

            regards, tom lane