ssh-copy-id command examples

ssh-copy-id — use locally available keys to authorise logins on a remote machine

If you have already installed keys from one system on a lot of remote hosts, and you then create a new key, on a new client machine, say, it can be difficult to keep track of which systems on which you’ve installed the new key. One way of dealing with this is to load both the new key and old key(s) into your ssh-agent(1). Load the new key first, without the -c option, then load one or more old keys into the agent, possibly by ssh-ing to the client machine that has that old key, using the -A option to allow agent forwarding:

user@newclient$ ssh-add
user@newclient$ ssh -A old.client
user@oldl$ ssh-add -c
... prompt for pass-phrase ...
user@old$ logoff
user@newclient$ ssh someserver

now, if the new key is installed on the server, you’ll be allowed in unprompted, whereas if you only have the old key(s) enabled, you’ll be asked for confirmation, which is your cue to log back out and run

user@newclient$ ssh-copy-id -i someserver

The reason you might want to specify the -i option in this case is to ensure that the comment on the installed key is the one from the .pub file, rather than just the filename that was loaded into you agent. It also ensures that only the id you intended is installed, rather than all the keys that you have in your ssh-agent(1). Of course, you can specify another id, or use the contents of the ssh-agent(1) as you prefer.

Having mentioned ssh-add(1)’s -c option, you might consider using this whenever using agent forwarding to avoid your key being hijacked, but it is much better to instead use ssh(1)’s ProxyCommand and -W option, to bounce through remote servers while always doing direct end-to-end authentication. This way the middle hop(s) don’t get access to your ssh-agent(1). A web search for ‘ssh proxycommand nc’ should prove enlightening (N.B. the modern approach is to use the -W option, rather than nc(1)).