ssh wont log in

[Red Squirrel]

I've seen this on multiple occasions and on various distros. The only way is to physically go at the console and reboot the server.

Right now it's happening on my live server. It accepts my logon, but it just stays stuck at the "last logon by" line. Anyone ever see this before? Is there a fix other then rebooting?

 

[M00T]

Figure out what it's doing:

ssh -vvv

 

[Crusty]

I usually get delays with ssh logins when the machine you are logging into has a high load on it, maybe you're hitting a timeout value before the login can complete?

 

[skyking]

Edit sshd_config to enable verbose logging and see what is happening.

 

[Red Squirrel]

Managed to get in after a while, so I'll edit that file and see what the log says next time.

 

[Nothinman]

If it got in after a while it's most likely a DNS lookup taking a while to timeout.

 

[Fallen Kell]

In some distributions/unixes, case in point Solaris, by default, unless you edit the /etc/default/login file, and comment out the line which says "CONSOLE=/dev/console" then root will not be able to login unless the device is the console. This could be what you are running into. Or you might have "PermitRootLogin no" set in the /etc/ssh/sshd_config file....

 

[Red Squirrel]

well it logged in before with same credentials, and it did eventually log in now and I've seen this on multiple servers not just this particular one (Was first time on this one). I don't log in as root as I have that disabled for security.

It seems to happen when the load is high, like if there's a backup job running. But a login should not really take that much resources to be affected by that. Just need to hash the password I gave and match it up with the account name and existing hash.

If it's a DNS lookup issue is there a way to suppress that? I remember having an issue like this in Apache where it would take about 3 hours to shut down the server because the DNS service shuts down before apache, then apache for whatever reason looks up every single virtualhost name on shutdown taking 10-20 minutes per vhost.

 

[Red Squirrel]

Started happening again, I have a web app that shows the load and something is eating up the cpu, load is at like 5+. Is there a way to stop ssh from doing dns lookups? That's probably what's causing it. I just want to be able to log in so I can figure out WTF is causing it to eat this much resources.

 

[Nothinman]

Why is DNS not working? That would probably be the best thing to fix.

 

[child of wonder]

UseDNS no

 

[Red Squirrel]

Originally posted by: Nothinman
Why is DNS not working? That would probably be the best thing to fix.

If the load is 5, nothing will work, since the server is simply too overtaxed. I never found out what caused the load to skyrocket though as I was not able to login.

 

[Nothinman]

If the load is 5, nothing will work, since the server is simply too overtaxed. I never found out what caused the load to skyrocket though as I was not able to login.

Not true, load average doesn't mean much at all. Unless something's changed recently even processes waiting on I/O contribute to load so you can have 30 processes waiting on the network so that you have a load avg of 30 and everything local on the box will perform fine because it's just the network that's being taxed.

I would guess that you had some sort of disk I/O problem and that's why you couldn't login. Things like writing to logs, spawning a new shell, reading shell config scripts, etc require a lot more resources than a DNS lookup.