We have tried sendgarbage both ways and it did not affect the timeout. In the TCP/IP Keepalive options, the "interval 20" is 20 seconds, not 20 minutes, so that is not too likely a suspect unless there is a 60 consecutive intervals limit buried someplace else.
Regards, Richard Schuh -----Original Message----- From: The IBM z/VM Operating System [mailto:[EMAIL PROTECTED] On Behalf Of Alan Altmark Sent: Wednesday, June 13, 2007 2:06 PM To: IBMVM@LISTSERV.UARK.EDU Subject: Re: TCPNJE On Wednesday, 06/13/2007 at 10:18 MST, "Schuh, Richard" <[EMAIL PROTECTED]> wrote: > We have success of sorts. The link can now recover from the socket errors it > experiences approximately every 20 minutes when it is not kept busy. The MVS > folks are questioning why these timeouts occur. I see the same time-outs on a > VM to VM link. The network people tell me that MVS is getting a?getpeername? > socket error. > > Is this something that is normal? If not, how do I fix it? The 20 minute timeout is likely because of your keepalive interval. Try changing it to SENDGARBAGE FALSE. Some IP implementations aren't happy with ill-formed TCP packets and z/OS or an intervening firewall may well believe it to be an attack of some sort and close the connection. Packet traces are, of course, de rigeur to solve such problems. Alan Altmark z/VM Development IBM Endicott
