Although it's not really my system yet to change at will, I did build the new TCPIP parms when the Linux guest CTC links were defined. It's true I haven't verified that they weren't changed after I sent them to the person who installed them. But as I wrote them, I specified an MTU size of 1492 on the VM side of the CTC link - just because (a) that's what the redbook used, and (b) I figured anything under 1500 would not cause fragmentation in the rest of the network. We also specify 1492 to the Linux code at boot.
I took another look at the VM ip parms. There are 2 other links defined in that ip stack - one to a guest OS/390 system also via VCTC, which specifies an MTU of 2000 on the VM side. That link works fine in every way, so far as I can tell (though most telnet traffic is 3270 emulation). The other is from the VM system to the outside world, an LCS link with an MTU of 1492. It took me several hours, once I started digging, to convince myself that command input was being accepted but no output was coming back. What finally did it was watching an io trace from VM. I could see the i/o for every character of telnet input I typed, up to and including the return at the end of the line, then nothing. The coup de grace, convincing even those who don't know VM TRACE, was having them telnet in from a local CMS guest, then watch me issue the "kill -9" for their pid from my "dead" Windows telnet session. What really bugs me is that the user and password prompt at logon appear normally - like as soon as the shell process is started stdout just goes to the bit bucket. So do you, collectively or individually, have any particular recommendation for a fix or experiment to run tomorrow when I get to work (short of applying the TCPIP maint)? Drop the MTU size to something below 1492 and see what happens? Thanks again to all of you who have responded for your thoughts on this. Steve On Mon, 25 Nov 2002, John Summerfield wrote: > On Mon, 25 Nov 2002 05:04, you wrote: > > That's assuming the MTU for the VM/TCPIP stack is >= 1492, which may not be > > the case (but most likely is the case). Rather than hit and miss, checking > > what it is and adjusting from there is important. > > Isn't the relationship supposed to me MTU <= MRU where MTU and MRU are > values at the opposite ends of the link? > > I can send as much as I like so long as it's no more than you expect > to receive. -- Steve Marak -- [EMAIL PROTECTED]
