I have got a more than 1000 ssh connection to one server, how to kill all without restarting the service.

kill -9 PID is difficult to handle with this issue.

ssh connections,

lsof -i :22 | wc -l
1241

Is there any way to control ssh sessions like, by finding and kill the inactive sessions.

Actually, that's 2 different questions.
1) Kill all without restarting the service: "killall ssh"
The service is named 'sshd' so it should still run. (And of course, test this first!)

2) Kill the inactive sessions
This is different and a bit more tricky.
But the first thought is, why are inactive sessions not dropped? This you should look into, manually killing inactive sessions is just a work-around until real problem is solved.
To find the inactive sessions you'll have to define "inactive", a session idle for how long?
Then use the command "w" and check the IDLE column. It will tell you username & TTY, you will have to use for example "ps -e" with grep to get the PID.
You could then create a script that finds & kills all sessions that have been idle for X minutes.

1 members found this post helpful.

Good advice.

However, it becomes really fun trying to decide what really is "idle", is a session running "top" for 2 hours idle or is it someone monitoring something?

Not to mention, is it idle when a user has a forwarded X application?

Is it idle if no tty is associated with the session?

Just kill em all and let god sort them out.

1 members found this post helpful.

In /etc/ssh/sshd_config you could insert

ClientAliveInterval 300
ClientAliveCountMax 3


That will give you an timeout of 15 minutes = 300 sec x 3 checks

That identifies network outages, or remote client failures.

It is possible that is what is happening though (it is also supposed to be the default, but individual distribution variations could have it turned off).