Dropped/hung TCP sessions on both NIC's, Compaq server with Debian
I'm running Debian unstable on an old Compaq dual proc P3 Proliant server box. I'm using the stock Debian SMP 2.6.something kernel.
Issue: Both the built-in 10/100 NIC (intel chipset I believe) and the second (PCI) NIC (3Com 3C9xx) are behaving badly. It's really odd.
1) Boot server.
2) Ping works to both NIC's.
3) SSH to box.
4) SSH connects and stays connected. But if you put in a command that lists more than about 1 line of text, the ssh session hangs ... does not drop, but will progress at about 1 line of output a minute if that.
5) Fresh ssh to the box will log in ... same behavior.
6) Continuous ping to the box does not drop a single packet during all this.
7) Have tried 2 known good cables, 3 known good ethernet switches.
The box worked fine when I did a netinstall on my home network (the server is at a parochial school.) The issue may have been going on for quite some time ... today I was over there to image some PC's and get the server into production; that's when I discovered the issue.
The issue as I saw it earlier today; samba connections to a share with the PC Ghost images would work for 3-15 minutes, then just stop. Rebooting the server would allow new connections to work (until those too failed). ifdown/ifup would NOT fix it ... only a reboot. When in failure mode, the box could not ping anything outside itself, but reported both interfaces up and running, and ifdown/ifup would result in a message saying that autonegotiation had succeeded.
Since ssh is useless for practical purposes, I'm looking for ideas/things to try/commands to run to help diagnose the issue. I'm not at all deep on Linux, but am a Cisco network guy and know my way around hardware etc. quite well.