We've been experiencing a strange problem that is baffling our mail and network gurus. Our listserv server is hosted on a linux/390 RH 7.2 image. Aside from some hiccups when the server was first built, listserv has just worked. Recently, however, I moved the listserv image from the RH 2.4.9-38 kernel to 2.4.19 with the IBM patches applied and moved from IUCV connections to a guestlan architecture.
We immediately began seeing undelivered mail to a few specific sites pile up in the mqueue. Most of the listserv traffic was being delivered but a few sites returned connection refused or connection timed out. These sites, incidentally, also feature higher security then the others. The DNS servers at our site and at the recipient's sites both had the correct info, so it doesn't appear to be a lookup issue. Sending mail manually from the command line (and telnetting to 25) also elicited the same responses. We moved listserv back to the IUCV connection and original ip but the behavior remained the same. In the course of testing, we found that some RH 9 boxes that we have also refuse to take any mail from the server. A quick glance at the logs and sniff of the traffic shows that the connection is rejected before any mail negotiation takes place, as if the packets are malformed or unacceptable in some way. We can traceroute and ping successfully, however. Does this ring any bells for anyone? I would backlevel the server to 2.4.9-38 (although I'm not convinced that the kernel is the problem), but I can't seem to find the rpm any longer. TIA, Michael Lambert
