Reposting: TCP connections getting randomly cut off. Even loopback connections.

Has ANYBODY seen ANYTHING like this happen?

A week ago, I posted a question about a situation we've literally never seen before, in well over a decade of working with TCP connections. It appears to be happening with both the proprietary protocol of our CRM product, and with secured Telnet sessions, and it happens randomly. The only thing that the two appear to have in common is that they both maintain persistent connections.

Starting sometime (if I remember right) in January, we've had a customer with some weird things happening to TCP connections.

They will disconnect, seemingly at random, for no apparent reason, and often (particularly with terminal sessions via Telnet) with the job failing to notice the disconnection.

I have, on several occasions, had two terminal sessions going with the system in question, and one will lose its connection, while the other will not, and the job for the lost connection will just sit there in WRKACTJOB, apparently completely unaware that it was no longer communicating with the terminal emulator, and it will often take several *minutes* to terminate it, even with OPTION(*IMMED).

And with the CRM product's propriatary protocol, it even happens (and indeed, was first noticed) in loopback connections: the web server for the product, running on the same physical box, connects to the proprietary server via an explicit IP address of 127.0.0.1.

--
JHHL