 |
The Question is:
In our process automation control systems , several servers are communicating .
A cluster of two new ES40 servers are also communicating with other servers.
Process automation tasks are distributed on two ES40's and are using C socket
interface when communicatihg with ather tasks on different servers. The cluster
of two ES40's has a cluster IP address . Every ES40 has its own IP address. Our
goal is:
a) To run the process tasks distrubuted to two ES40's .
b) To make possible the other servers to communicate with its partner process
using cluster IP,
independently of ES40 server on which it is running.
c) To move the processes to one of the machines in some cases ( such as
maintenance, failure etc.)
without effecting the other servers communicating with these ES40 servers .
In the present situation , when we start the task on one of the nodes , the
client task can connect,
but if we move a process to the other ES40 server, its client task can not be
connected using
cluster Ip giving "connection refused message .
Before making "listen system call" we pass the "Inaddr_any" parameter to the
bind system call.
Requesting a way to overcome that problem.
Best Regards.
The Answer is :
This task appears to fit what DNS and similar directory services
packages are designed for. Failing the applicability of a standard
directory services package such as DNS, you will want to create your
own application-specific directory services package.
Within an OpenVMS Cluster, the metric server and the cluster alias
will be of interest. Also of interest will be the inetd daemon,
and its ability to automatically start the necessary server(s).
Also portmapper, obviously.
There is insufficient information to determine the specific cause
of the IP socket error(s) reported. If you wish to pursue the
resolution of the error(s), you will want to contact the Compaq
Customer Support Center, and you should expect to be asked for the
IP stack and version (TCP/IP Services V5.0A with ECO or later is
preferable), the specifics of the source code and the error, and
(if possible) a reproducer.
|
