Steve, The first thing I'd do is to check system performance group settings. Is your session manager task running in a performance group at least as high as VTAM and TCPIP? If the system is 100% CPU busy and either one or all three tasks (VTAM, TCPIP, SessMgr) get starved for CPU cycles and cannot respond to each other in a timely fashion, I'd say, all bets are off and you'll begin to see strange things happen.
Regards, Ulrich Krueger -----Original Message----- From: IBM Mainframe Discussion List [mailto:[email protected]] On Behalf Of Thompson, Steve Sent: Thursday, June 18, 2009 07:29 To: [email protected] Subject: TCP Settings for KeepAlive I've run into a problem where I have TN3270 sessions attached to a VTAM session manager. The problem is this: When the system, where the VTAM session manager is running, is at 100% for extended periods (5 minutes or more), AND, you have a TSO session on the system where the session manager is, AND that TSO session attempts to run a "long running task" (Endevor update to promote source is an example), the TSO session will continue to run, BUT the TN3270 session will be timed-out and dropped. In this case the TN3270 is QWS3270 Plus. In tracing this, we see that QWS is "posted" with a lost connection. However, QWS is running the TCP keep alive at about a 5 minute interval. And the last thing that QWS shows is it sent a buffer to the host and is now waiting for a response when it gets this lost connection posting. I have been looking at the z/OS Comm Server manuals (even searching them) and I don't see anything that needs to be changed in the configuration. I have looked at various APARs (particularly the ones about IMS and additional keepalive support from 1.7 and later). But I don't see anything that seems to apply. The other thing about this that is interesting, those running Hummingbird don't seem to have this drop problem. And this is something that I can recreate at will. Note: TSO is not timing out. This is specifically a TCP connection termination issue that is timing related. It takes the z/OS system (1.8) going to 100% busy for 5 minutes or longer. And, while this is happening, I can be directly connected to other MVS or VM systems and they are not affected. So anyone have any idea where I need to look in the config parms for TCP (or VTAM?). Regards, Steve Thompson ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
