I have three machines networked to my desktop which run a bunch of simulations in parallel. As they're running, I connect to them via SSH and screen to keep an eye on the runs and look at the output. They stay usually connected for days at a time. The SSH servers and client are running Fedora 14.

Yesterday one of my coworkers accidentally yanked the plug on one of the servers while it was running. When I powered it up again, I started getting some odd connection problems. I couldn't connect to it via SSH initially because I got the Remote Host Identification Changed (RSA host key changed) error. I deleted the key in .ssh/known_hosts, which allowed me to connect, but it denied my password. I then logged into that machine locally, restarted sshd, and removed .ssh/known_hosts again. Now I can log in via SSH without problems. However, the connection dies with a "Write failed: Broken pipe" error every few minutes (as opposed to the other two machines, which stay connected indefinitely).

So my questions are:
1.) why would a power loss affect the behavior of the SSH server?
2.) why do I keep getting broken pipes now?

Thanks!

A reboot causes services to be restarted, so config files are re-parsed. If there was an update to the ssh-server or changes made to the config file, they would not be made live until the server restarted. To fix the broken pipes issue, open your ssh configuration file (usually /etc/ssh/sshd_config) and check for ClientAliveInterval and ClientAliveCountMax. I suspect these items because it sounds like the client machine is not responding to keepalive requests. You can check the man page for sshd_config for more details on those and other directives that may be pertinent. You could also try simply copying over the sshd_config from one of the other servers.

Off the top of my head, try "ifconfig" and see if you're getting any packet errors or drops on your interface. The power drop might have munged it.

ifconfig shows no errors and no drops.

This apparently isn't just an SSH problem. Earlier today I connected through VNC, about an hour ago VNC failed to connect, and now it's working again.

I checked the SSH config files, and they are the same as the other machines, so nothing changed. Also, that doesn't seem to explain why it denies my password.

This is starting to get me concerned. Thanks for your input, let me know if you have other ideas.

I borrowed another ethernet card and popped it in there. That one works fine... all the connection problems are gone. I didn't think it was a hardware problem because I figured the RSA key wouldn't change if something was broken, it just wouldn't connect at all.

I guess I was wrong.

This is why I can't have nice things.