Re: PGSTAT : COnnection Refused Error (pgsql 7.2-2 on cygwin

Tim,

On Tue, Mar 05, 2002 at 02:48:02PM +0000, Tim Finch, FosterFinch Ltd wrote:
> At 07:50 04/03/2002 -0500, you wrote:
> >3. Try to strace the problem:
>
> I did your 'unlink and copy' procedure then
>
> strace --output=strace.output postmaster -i -D /usr/share/postgresql/data
>
> (not even in background, and all output to stdout)
>
> err. it RAN! I managed to get pgAdmin to connect to it, and even got psql
> working (which you may recall was a previous thread issue with it
> crashing).

Unfortunately, using strace can "fix" problems -- especially timing
related ones.  This is due to the fact that strace-ing slows down the
traced process.

> Great! So how do you stop strace?? I ended up killing it from
> another bash screen.

Just killing it fine.

> >Any thing obvious in postmaster.log?
>
> Well here it (below) is at '-d 4', but without strace.output (when I did
> the strace on postmaster it worked so no error messages in log)
>
> [snip]
>
> PGSTAT: connect(2): Connection refused
> DEBUG:  proc_exit(1)
> DEBUG:  shmem_exit(1)
> DEBUG:  exit(1)

I grep-ed the code as follows:

    $ find . -name '*.[ch]' | xargs fgrep 'PGSTAT: connect'
    ./src/backend/postmaster/pgstat.c:              perror("PGSTAT: connect(2)");

Hence, postmaster is crapping out at the "if" below:

/* ----------
 * pgstat_init() -
 *
 *  Called from postmaster at startup. Create the resources required
 *  by the statistics collector process.
 * ----------
 */
int
pgstat_init(void)
{
..
    /*
     * Connect the socket to its own address.  This saves a few cycles
     * by not having to respecify the target address on every send. This
     * also provides a kernel-level check that only packets from this same
     * address will be received.
     */
    if (connect(pgStatSock, (struct sockaddr *) & pgStatAddr, alen) < 0)
    {
        perror("PGSTAT: connect(2)");
        close(pgStatSock);
        pgStatSock = -1;
        return -1;
    }
..
}

Are you configuring postmaster to collect stats?

Sorry, but I cannot reproduce your problem so I cannot debug it further.

Jason