monty@skipper.dfrf.nasa.gov (Terry Montgomery ) (10/12/89)
My corner of the local (Ether)network is irrational. I'm getting that way myself :-(. In a lab, we have a Sun 386i workstation and 4 PCs running SUN's PCNFS software (version 3.0.1) on 3C503 boards. All are connected to a length of cheapernet which is connected to the rest of our network via a multi-port repeater. Each unit behaves well with the rest of the net. The PCs and workstation can telnet to computers throughout the site. Computers throughout the site can telnet to the workstation. Ping (and/or nfsping) works. The PCs can accept yellowpages service and can mount virtual disks on a remote workstation through two repeaters and a bridge. BUT the PCs and the workstation are mutually invisible. NO telnet, ping, nfsping... The workstation is supposed to be a fileserver for the PCs! What I think that I see on the net (using a borrowed LanAlyzer) at the start of a telnet session is a PC doing an arp broadcast request and the workstation doing an arp reply (which is apparently ignored). The arp request/reply sequence is repeated several times before telnet on the PC declares failure. I have tried to configure the PCs to remove NFS and yellowpages dependencies (simplify! simplify!). Not fixed. I replaced a 3C503 board with a Excelan board and PC-NFS with Excelan HostAccess(?) software in one of the PCs. That PC can now telnet to the Sun workstation. The workstation can ftp to the PC when the PC is running as an ftp server. The PCs using NFS cannot ftp to the reconfigured PC. So, the problem is now (probably) reduced to questions about PC-NFS and 3C503 cards. Identical 3C503 cards and identical PCNFS software is working perfectly in a room on the other side of the building. And yes, they assisted in the PC-NFS installation in the lab. The 3C boards run echo server diagnostics with no problem. I have replaced the transceiver between the cheapernet and the 386i. What is going on? All problems (except timing!) that I consider are incompatible with the fact that the systems, boards, transceivers, and cables do function for remote (100s of feet) communications. I assume that 3Com can handle a fast response. Can YOU suggest why a network might function well over large distances, but fail across the room? Terry Montgomery monty@elxsi.dfrf.nasa.com No disclaimer authorized.