Tom Lane wrote:
> "Siddharth Rath (sirath)" <sirath@cisco.com> writes:
>
>>Surprisingly .s.PGSQL.5432 file is not getting created in /tmp directory ( I believe this should be created).
>
>
> No, it shouldn't, because that is associated with the creation of the
> Unix socket file, which happens after creation of the TCP socket.
>
> Given that you're seeing a kernel error on attempt to create the TCP
> socket, I'm convinced that you're wasting your time thinking about
> Postgres-internal conditions. This is a kernel-level issue and you
> need to be inquiring into what state the kernel is in that is making
> it reject the bind().
>
> regards, tom lane
You might try to set up a simple web server like thttpd or simpler on
port 5432 and that way you can troubleshoot opening the socket listener
is a more simple environment with possibly some different error
messages. Also you can run a simple perl/python/ruby/java script (like
this http://myclass.missouri.edu/~atb6x2/NETSERV/NS_serverpl.html) to
(try to) open a socket listener on 5432 and have it just sit there.
If you do that the output from strace would be shorter than trying to
start up pg and you might get an idea where the socket problem is.