SSH to two addresses, use the one that connects first

Question

I have a home computer (let’s call it franklin because that’s what I call it) that I often ssh into from my work laptop. When I’m at home, I ssh to franklin.local, and when I’m at work or anywhere else, I ssh to remote.address.of.franklin.

I want to have a function in my profile that always connects in the correct way for the situation. My network conditions often change (wired vs. wireless, different SSIDs, etc), so I’d prefer not to do it by detecting the state of the network, but rather:

1. If franklin is available locally, connect locally,
2. else, connect remotely

Here’s what I use now:

function franklin () {
    ping -c 5 "franklin.local" > /dev/null
    if [[ $? != 0 ]]
        then
        echo "franklin not available on local, so trying remote connection"
        ssh -p 22 remote.address.of.franklin
    else
        echo "franklin available on local network, so trying local connection"
        ssh username@franklin.local
    fi
}

The problem is, the ping takes time. My question is: can I execute both ssh commands simultaneously, and then automatically use whichever one connects first?

Answer 1

What about trying a simple OR from your shell?

ssh username@franklin.local || ssh -p 22 remote.address.of.franklin

I am not really familiar with ZSH, but I guess the evaluation logic would still be lazy, meaning that the second part is only evaluated if the first part fails. Of course the first command may stall for some time trying to figure out if franklin.local is available.

You could assign that command to an alias to shorten it up (like you did with the function).

Answer 2

Turns out I was closer to the solution than I thought. I had a faulty assumption about how the address lookup part of ping works. I thought that if I had a network hiccup, and franklin.local didn’t return the pings, then it would fail and connect remotely instead. Turns out, if franklin.local is on the network, ping -c 1 franklin.local returns 0, and if not, it returns 68, both pretty immediately. So the only change is the count of the ping from 5 to 1 (I've also tweaked the order of the script to make it make more sense):

function franklin
{
    ping -c 1 franklin.local &> /dev/null
    if [[ $? == 0 ]]; then
        echo "Connecting locally."
        ssh "username@franklin.local"
    else
        echo "Connecting remotely."
        ssh "remote.address.of.franklin"
    fi
}

This is significantly faster than how I had it before, and not appreciably slower than just using the appropriate ssh command directly.

Answer 3

I use nc instead of ping:

franklin() {
    local host=franklin.local
    if nc -w1 $host 22 &>/dev/null; then
        echo "connecting locally"
    else
        echo "connecting remotely"
        host=remote.address.of.franklin
    fi
    ssh username@$host "$@"
}

I added "$@" to the ssh call. That let's you quickly execute remote commands like

franklin uptime