gxadmin gunicorn handler-restart
currently restarts one gunicorn instance after another, which leads to very long restart times (and therefore very long playbook runs on the headnodes)
To avoid this, we can read the running instaces into an array, split them in half and then restart first half and second half once the first half is serving requests.
from