There was no problem to log in the remote computer, which never needed a second try. However, the connection frequently got killed by the remote server, sometime when I logged in for 1 minute, sometime even when I just logged in.

Could any one tell me what or how to locate where the problem is?

Try setting the following in your /etc/ssh/sshd_config file and restarting SSH on the remote server.

TCPKeepAlive yes

Thanks any way.

Is your connection within the same LAN, or does it go over a router to some other LAN? If the latter, is it local or over the internet? Any chance you have two different paths to the same remote computer?

Can you see the connection status in the output of "tcpdump" on both ends at the same time before the connection drops? I am curious if the source port number matches and whether the addresses are translated.

Thank you so much.

I first ssh a remote pc over internet, this connection never has problem. Then within this ssh, I ssh another pc(192.168.88.5) over LAN, it is local, and the problem occurs. Sorry that I am not familar with network protocol. However, I　tried ping 192.168.88.5, and 100% packages are successfully tranmited.

Should I track the problem use "tcpdump"? My remote PC uses UBUNTU, it has tcpdump command.The pc(192.168.88.5) uses CENTOS, and it did have installed tcpdump, however, it is strange that it does not have the tcpdump command but the tcptraceroute command.

Can you explain what you mean by "within this ssh"? I can think of more than one thing for this, so I need to narrow it down. Can you show both ssh command lines in full?

I don't know if this will be needed, yet, until I know the answer for above.

first, ssh the remote PC by internet
>ssh 166.111.XX.XXX

after log in, ssh the pc in LAN in the same terminal
>ssh 192.168.88.5

which means the remote PC connects both the internet(by ip 166.111.XX.XXX) and the local LAN (by ip 192.168.88.6). I need to log in the remote PC in order to access the PC in the local LAN(ip is 192.168.88.5)

thanks.

What you have here are 2 logins. One from host A to host B. One from host B to host C, done from the session started by the previous.

Now which connection is getting killed by which remote?

from host B to host C (i.e. the ssh connection in the LAN)

thank you so much.

So when that connection gets dropped, the connection from A to B is still up, and you can still type commands to host B.

Can you get to the physical console for host B at some point in time to debug from there by trying to connect from B to C there?

Yes.

Your suggestion is good, unluckily, I am currently in anther city, so can not access the physical console for host B. I can access the the physical console after a few days, will try the suggestion and report the results.

Thank you so much, Skaperen.

Can you login as root on C, run screen, and do a tcpdump in the screen session to watch and save the packets ... and likewise on B ... to see what might be causing the connection drop?

Now, I can access the physical machine of remote PC B. Even I directly use PC B to ssh PC C, the problem occurs. After a few tries, I am sure that the problem is caused by the ssh-client of this PC.

I rebooted the PC, ran Knoppix (which is a live Linux CD), and all problems disappeared.

Thank all of you above.