Socket IO fails with BROKEN PIPE

[Willem Grooters]

An application on VMS creates a socket to an application on TRU64 (or another Unix), sends data over that socket (~ 1490 bytes), and next will send another message (~ 65 bytes) and after that, a larger number of similar messages, in size limited to about 80- byets each, and in the end it reads back an aknowledgement message.
On the same port, messages are transferred from the Unix application to a TCPIP-service on VMS that ends existance once the message has been processed - this port id therefore bi-directional.

This works fine but for one site: The first message is transfered without error but sending the next fails with errno=32, errtxt = "broken pipe".
I couldn't find any information on this error in the VMS manuals, and searching HP site for "EPIPE" revealed a number of documents on HP-UX and an article on the differences between TRU64 and HP-UX.
Since this happens on that site only, I guess it is something on the Unix side; traffic the other way seems to have no problem.
My questions are:
* What could cause the error
* What can (and should) be done at the remote site to prevent the problem?

(VMS version 7.dunno, TCPIP 5.dunno either)
(TRU64? version: unknown)

[Jim_McKinney]

> * What could cause the error

The remote end closed the socket.

> * What can (and should) be done at the remote site to prevent the problem?

If there is an intervening firewall, insure it is not dropping the connection. If you rule that out, then look for a programming error on the remote end.

Using a tool such as TCPDUMP while a connection is in progress may prove enlightening.

[Wim Van den Wyngaert]

Jan,

Post a tcptrace/prot=tcp/fu.

Is there internet between the 2 nodes ?

We recently had a problem of unknown nature and switching to another phone (inbelpunt) solved the problem.

Wim

[Willem Grooters]

Bad habit by previous programmer not to supply the right values. Chnaged code and found reason - in the other program.