Comment 10 for bug 1644530

On Zesty (still without the pidfile in the service) it seems to work fine (Version 1:1.3.2-1).

Test:
2x zesty KVM guests
sudo apt-get install keepalived
Set up as above, but with ens3 (virtual ethenet device).

I think the name of backup/master conf names are interchanged in the example above, but it works fine.

0) pre restart
  Process: 2173 ExecStart=/usr/sbin/keepalived $DAEMON_ARGS
 Main PID: 2183 (keepalived)
           ├─2183 /usr/sbin/keepalived
           ├─2184 /usr/sbin/keepalived
           └─2185 /usr/sbin/keepalived

1) First restart
  Process: 2393 ExecStart=/usr/sbin/keepalived $DAEMON_ARGS
 Main PID: 2404 (keepalived)
           ├─2404 /usr/sbin/keepalived
           ├─2408 /usr/sbin/keepalived
           └─2409 /usr/sbin/keepalived

I see the backup server properly detect, transtion to take over and back to backup.
I was restarting on the master above, now trying on backup.

0) pre restart
  Process: 2137 ExecStart=/usr/sbin/keepalived $DAEMON_ARGS
 Main PID: 2147 (keepalived)
           ├─2147 /usr/sbin/keepalived
           ├─2148 /usr/sbin/keepalived
           └─2149 /usr/sbin/keepalived

1) First restart
  Process: 2765 ExecStart=/usr/sbin/keepalived $DAEMON_ARGS
 Main PID: 2776 (keepalived)
           ├─2776 /usr/sbin/keepalived
           ├─2777 /usr/sbin/keepalived
           └─2778 /usr/sbin/keepalived

Note: since they did not add the PIDFILE we might want/need to understand what was done to fix it here (to backport that as SRU)
Next: I'm trying to reproduce on Xenial as reported