Multipathd service hangs during startup and gets killed by systemd when there are non responding paths

Issue

• Starting or restarting multipathd fails with systemd killing the process after approximately 3 minutes.

• The commands systemctl start multipathd and systemctl restart multipathd hang. After approximately 3 minutes the following output reports their failure:

  Job for multipathd.service failed because a timeout was exceeded.
  See "systemctl status multipathd.service" and "journalctl -xe" for details.
  

• Messages similar to the following appear in the system logs:

  Mar 27 10:15:42 hostname systemd[1]: Starting Device-Mapper Multipath Device Controller...
  Mar 27 10:15:42 hostname multipathd[308729]: --------start up--------
  Mar 27 10:15:42 hostname multipathd[308729]: read /etc/multipath.conf
  Mar 27 10:15:42 hostname multipathd[308729]: path checkers start up
  Mar 27 10:17:12 hostname systemd[1]: multipathd.service: Start operation timed out. Terminating.
  Mar 27 10:18:42 hostname systemd[1]: multipathd.service: State 'stop-sigterm' timed out. Killing.
  Mar 27 10:18:42 hostname systemd[1]: multipathd.service: Killing process 308729 (multipathd) with signal SIGKILL.
  Mar 27 10:18:55 hostname systemd[1]: multipathd.service: Main process exited, code=killed, status=9/KILL
  Mar 27 10:18:55 hostname systemd[1]: multipathd.service: Failed with result 'timeout'.
  Mar 27 10:18:55 hostname systemd[1]: Failed to start Device-Mapper Multipath Device Controller.
  

• Multipath commands are also hanging on the system.

• Depending on the state of multipath maps, IO operations to multipath maps may not be hanging: If the hanging paths have already been failed by earlier errors or timeouts, then IO to the multipath maps will not be hanging.