I have inherited a solaris 10 server. I went to log into it the other day over the console connection, and after sucessfully entering the password, I was booted back to the login prompt with no errors visible.
If I enter an incorect password, the console displays 'incorrect password' and kicks me back to the prompt. A correct password does not display this error but still kicks back to the login prompt. A hard reboot (hitting the power button) does not resolve the problem, and there are no errors apparent when the machine is booting.
I did not set up the server and i have no idea what the guy before me did to it, he has now left. Any thoughts on what could be wrong?
I considered a full HDD but could not recreate the problem on a virtual machine? I would also expect more errors on the boot sequence if the disk were indeed full.
Boot off CD or a jumpstart server and mount your root disk. Check your passwd file entry. It's possible that the shell definition is incorrect. Check your messages file. Check that your defined home directory exists. Also, check if /etc/nologin exists. Although, that won't keep root from logging in from the console. Assuming that you are logging in as root.
Thanks for the reply, just need to figure out how to get to the OK> prompt over a telnet connection to the console, I think it best to directly connect these things up.
And yes, my predecessor seemed to have committed many enourmous admin faux pas, only a root account, no quotas on zfs mounts, etc etc.
From experience, the server has been fine (in one manner of speaking at least), in fact i was able to log in about a month ago just fine, and i think the people with access to the box is very limited. Very strange.
Sorry forgot to mention that, its a SPARC processor, im not on site at the moment so I cant tell you much more for now. I am assured there is 'plenty' of disk space.
The server appears to be a Sun T2000, but im going on memory, and what a google image search shows up!
Assuming that you are able to logon to the sc, you can run the braek command to send a break to the OS. Be aware that you may have some clean-up afterwards. You might have to run anfsk on a filesystem etc.
Single user mode gets me in ok, and i can mount all zfs drives etc. I noticed the authlog is about 60Mb+ and logadm hasnt run for about 15 months, and no one noticed.... (disclaimer, i only came to this a few weeks ago).
The authlog shows a load of cron auth errors for the root cron.
I created a test user in single user mode and set up the password, rebooted and tried to login as the new user, this failed, same problem as the root user, just get back to the login prompt with no error message.
There is also plenty of space on the drive, so its not a disk full problem. The hostname.[interface] file seems to have vanished, im sure it was there before, which explains why the server now refers to itself as 'unknown'.
Im beginning to think there may be a problem with the login scripts? does this sound viable? Any other ideas are welcomed as im stumped!
OK so i got to the root of the issue. Checking the (now enourmous)messages log gave a line similar to 'Could not open connection to ttyname' each time i tried to log in.
It turns out, that the PAM module didnt like the server name change (when the files mysteriously disappeared) and was not allowing logins. Once i replaced /etc/nodename and /etc/hostname.XXX i was able to log in properly.