startSSHAgent bash function - checks for SSH_AGENT_PID in env, then for ssh-agent PID (pgrep), if not found, starts new ssh-agent, adds your keys

startSSHAgent.sh

startSSHAgent() {
  if [[ -z "$SSH_AGENT_PID" ]]; then
      if [[ $(pgrep ssh-agent) ]]; then
          export SSH_AGENT_PID=$(pgrep ssh-agent)
          echo "Found existing ssh-agent PID, SSH_AGENT_PID=${SSH_AGENT_PID}"
      else
        echo "Starting fresh ssh agent"
        eval `ssh-agent`
      fi
  fi
  
  ssh-add ~/.ssh/my-private-key1
  ssh-add ~/.ssh/my-private-key2
  #...
  ssh-add ~/.ssh/my-private-keyN
  kill_old_ssh_agents #see https://gist.github.com/gubatron/2d97b31b0621c459f8b5ee8665c9f7b9
}

I have this one for some years now;

# first check if we have access to an ssh-forwarded agent.
let ssh_agent_done=0
if test -n "$SSH_AUTH_SOCK"; then
    if test -S "$SSH_AUTH_SOCK"; then
        let ssh_agent_done=1
    fi
fi

if test $ssh_agent_done -eq 0; then
    if test -f $HOME/.ssh/agent; then
        source $HOME/.ssh/agent > /dev/null
    else
        #initialize ssh-agent
        (umask 077 && ssh-agent -t 172800 > $HOME/.ssh/agent)
    fi

    if test -n "$SSH_AGENT_PID"; then # test if still alive
        if test -d /proc; then
            if ! grep -q ssh-agent /proc/$SSH_AGENT_PID/cmdline; then
                (umask 077 && ssh-agent -t 172800 > $HOME/.ssh/agent)
            fi
        else # no proc, but still unix. Probably Mac.
                # test if still alive
            if test -z "`ps $SSH_AGENT_PID | grep $SSH_AGENT_PID`"; then
                (umask 077 && ssh-agent -t 172800 > $HOME/.ssh/agent)
            fi
        fi
    fi
    source $HOME/.ssh/agent > /dev/null
fi
unset ssh_agent_done

hey @zander long time no see :)

is the umask 077 to verify that the current user has proper permissions to execute ssh-agent? (and not the world/group)
or that the ssh-agent process doesn't give permissions to others but the user launching it?

today I also learned about that -t <life> parameter, has there ever been a case in which you thought necessary to keep your sessions only 48 hours long? I can keep a terminal window open for a week easily

This approach of writing the ssh-agent output to a file > $HOME/.ssh/agent is also interesting. I just eval the output

Hi guba! Miss your patches at Flowee.org :)

The umask is probably not really needed, it was to make sure that the file is only readable by owner.

the limit of 48 hours is not on the life of the agent, it is on how long the agent keeps passwords. So you'd have to do a 'ssh-add' again after that. It is purely for security.

I just eval the output

That has the downside that if you logout your ssh-agent will keep running but you can't reach it anymore. Next time you login another agent will start. And you'll end up with more and more of them.
My writing file solution also means that if I login on the desktop and then later ssh to the same machine then I can reuse the already running ssh agent.

well I only eval if I can't find an existing ssh-agent, otherwise yes you end up with a bunch of ssh-agent processes

Ill check what's up with flowee.org and how I can help once I get a breather