At our site we have had lots of "issues" with users using HOD from
various geographic regions loosing sessions to TSO. When people were
crying that HOD was broken my point of view was HOD was probably
functioning as designed, something might not be right somewhere in the
network topology. Management finally got our outsourcer to work with
CISCO and some really highly skilled network folks to really look at
what was happening beyond the "I lost my TSO session" sort of problem
reporting.
After much investigation we have got most of the issues stabilized.
There were numerous contributing factors like back level router IOS
levels, network hardware that had been up for close to two years without
a reboot, and various network settings that CISCO recommended changing.
Lots of things have been fixed, rebooted, flashed, changed etc.
The one area left, in my opinion is some of the MVS TCP/IP settings.
We have the following settings in TCPPROF:
INACTIVE 0
TIMEMARK 900
SCANINTERVAL 120
and in SMFPRMxx:
JWT(0030) (This was (0060) until recently)
Back when JWT was an hour there seemed to be a clustering of reported
timeout values in end user reports. The problem was TSO sessions for
people half way around the world that were in use but idle would
reputedly drop after around 15 minutes of no activity, sometimes 30
minutes, or more frequently (in reliably repeatable testing) 45 minutes.
My contention was something in the TIMEMARK-SCANINTERVAL handshaking
with a TN3270 session was getting missed for users in an ISPF edit
session when they were not hitting enter for a long time.
Moving a cursor around, typing text etc doesn't do anything to say to
VTAM et al "I'm still active". Sending an AID like ENTER or a PF key is
needed for all the network layers to see traffic and reset timers like
TIMEMARK, JWT, etc.
The end users who were getting blown off TN3270 sessions at about 15
minutes were not be able to get in until the TSO timeout of an hour had
expired. Navigating all the international calling issues to get to a
help desk to get to an operator to cancel a TSO user id was more than
folks were either willing or able to do.
So, my question is, how many sites have a TIMEMARK less than JWT?
Knowing end user work habits there will sometimes be a lot of "think
time" before an AID is sent. Why time out the TCP/IP session before the
TSO session?
Oh, and I did RTFM and RTFA (read the "fine" archives), and I know IBM
recommends TIMEMARK to be many hours but there seem to be issues with
people believing that these values are good to use at our site. I'm
wanting to know what some of you all do so I can get some traction on
getting things changed.
Thanks!
Peter Duffy
Operations Support - Mainframe/AS400
NISSAN NORTH AMERICA
Information Systems Department
Gardena, CA, USA
310-771-6472
----------------------------------------------------------------------
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
