How to access my Linux Server when ssh is not working?
I'm doing a lot of fooling around with Linux at home--trying to prepare for job interviews. I am using an IBM ThinkCentre and I installed Fedora 8. When I do a "uname -a", I get the following:
Linux troyvnet 126.96.36.199-28.fc8 #1 SMP Sat Sep 20 09:32:58 EDT 2008 i686 i386 GNU/Linux.
I had a phone interview yesterday and he asked me how do I access my linux servers. I told him that I always use ssh (putty) from my desktop. He said they normally use an ssh client also. But, if the server is having issues and they can't use ssh or telnet to see what's going on, they would use the ethernet plug in the back of the boxes to gain access.
I just said oh, ok. But, I really did not know what he meant. Can anyone explain to me, in detail, the process to do this? I would like to test this out on my Linux server at home. I see one ethernet port in the back of the box; but how would I be connected to the network with that ethernet port and have to hook up some type of terminal or my laptop to actually see what's going on with the server? Do I need some type of ethernet adapter? Any ideas?
Either he meant an isolated second ethernet port that wasn't protected or he meant to say a "serial" port. I suspect the latter. There are devices that are like KVMs but switch serial ports to allow remote access to terminals without being on site with a crash cart.
You can allow logins on a particular serial port. Look at the manpage for "securetty" or "login.defs" and "inittab".
By the way, why do you use putty and not an ssh client on a Linux machine?;)
If I had to use windows, I would prefer using Cygwin/X. Then you have the same ssh client as in Linux, and you can run a system-config gui program remotely, even if the server runs in init level 3. However, you do need some of the Xorg libraries installed for that which you may be unwilling to do on a server.
jschiwal: Thank you very much for your input.
I guess I'm still a little unclear on what I need to do. This is all new to me and I'm trying to simulate this scenario for my home Linux environment. Do I need to buy a serial port switch/box/cable and what do I need to connect to what? Can you point me to some type of documentation discussing this?
As far as putty, our laptops are supplied by our job with specific software on them; because we use it to remotely access other systems also. But, I will look into Cygwin/X.
Of course, the server needs to be configured to allow logins in that port, using agetty or whatever (the relevant file here is /etc/inittab, which by the way has a man page that might come in handy). Then you can use almost anything to login to the serial port from the client machine: from a real hardware dumb terminal to the windows hyperterminal, and of course, a wide range of linux terminal emulators capable of emulating vt100/vt102 stuff.
We have some devices on remote locations. These devices insert commercials on cable networks and are connected to the headends internet switch. Also, the first device has a (pots) modem input, and there are daisy chain connections between all of the units in the rack. This is just an example. Regular servers don't have the same daisy chain inputs, but the principle is that if the network device goes out, you can still connect via a phone connection. You could even then work on the router, or switch remotely from a server.
There was an article in a Linux magazine earlier this year about the authors experience using bonding for redundancy. This may be more important to understand than a second login port. It entails bonding two interfaces to one IP address. Each interface is connected to a different redundant switch or router. The switches/routers eliminate the redundant routes, but will add them if the first connection goes down. A problem the author had is when the load increased on a switch, it wouldn't send a certain type of packet to the next switch and that switch would un-delete the redundant route causing a loop. The solution was to replace some switches with routers. Apparently there is a low limit on how many inputs you can use when using bonding. If you can find that article you may find it interesting.
You never can be certain what the next interviewer might ask. But you have the right idea, in trying out things yourself so you can rely on actual hands on experience. In a work setting, one probably would do things the same way for a long time. At home you can try out different setups. At home you can experiment, and learn for example, when an application server works best and when centrally mounting shared-static system directories works best.
I find I've been babbling on. Sorry. Good Luck!
|All times are GMT -5. The time now is 07:03 AM.|