Keeping an idle tcp connection up for more than 2 hours
Our HP TCP/IP Services for OpenVMS Alpha Version V5.6 - ECO 3 on an hp AlphaServer GS1280 7/1150 running OpenVMS V8.3 has been dropping it's connection to an interface engine after two hours of idleness. We don't now where the connection is being chopped but it does seem to happen after two hours of idleness.
It has been suggested that I can maintain this connection by adjusting a parameter in tcpip services that would exersize the connection periodically, like less than two hours. I've browsed the help and manual in tcpip services and googled it but I haven't found the correct setting I looking for. Can anybody help me here?
Re: Keeping an idle tcp connection up for more than 2 hours
TCP/IP Services does not include this feature. Accordingly, this has little or nothing with TCP/IP Services and almost certainly lurking within your local IP infrastructure or with your target.
Details of the connection would lend themselves to determining which particular keepalive is appropriate. Or whether there's a brute-force solution for the protocol.
I'd be looking most carefully at the interface engine, and at the network itself.
Details on the connection and the network and the identity of the interface engine involved here might help.
Re: Keeping an idle tcp connection up for more than 2 hours
Clark,
look at the 'tcpip_keepidle' mechanism.
Is a firewall involved here between those systems ? I've seen a case, where a firewall dropped an IDLE connection after 60 minutes when the OpenVMS (default) tcpip_keepidle was 14400 (i.e. 120 minutes). Lowering the value to 55 minutes on OpenVMS solved that problem.
Re: Keeping an idle tcp connection up for more than 2 hours
Clark,
as I assume this is a reproducable problem, consider to start a TCPDUMP trace on the OpenVMS system to capture all packets to/from the 'interface engine'.
