I deploy RHOSP13.
I am using ipmi as fencing mechanism for the underecloud's controllers and as a way for tripleo to power on and off the nodes.
I import the overcloud node with

openstack overcloud node import ~/instackenv.json

Excerpt from the instackenv.json:

       {
            "mac":[
                "00:00:77:88:00:02"
            ],
            "name":"controller01",
            "cpu":"4",
            "memory":"32768",
            "disk":"40",
            "arch":"x86_64",
            "pm_type":"ipmi",
            "pm_user":"admin",
            "pm_password":"redhat",
            "pm_addr":"192.168.104.1",
            "pm_port":"6234"
        }, 

The node can be powered on and off when needed during introspection and deployment.

In order for the controller node to be able to fence other controller node, i use a fencing.yaml that has been created with:

openstack overcloud generate fencing instackenv.json --output fencing.yaml

Excerpt from the file:

  EnableFencing: true
  FencingConfig:
    devices:
    - agent: fence_ipmilan
      host_mac: 00:00:77:88:00:01
      params:
        ipaddr: 192.168.104.1
        ipport: '6233'
        login: admin
        passwd: redhat

Now, when the undercloud is deployed, a "pcs status" show tht the fencing device are stopped:

 stonith-fence_ipmilan-000077880001     (stonith:fence_ipmilan):        Stopped
 stonith-fence_ipmilan-000077880002     (stonith:fence_ipmilan):        Stopped
 stonith-fence_ipmilan-000077880003     (stonith:fence_ipmilan):        Stopped

It show also this kind of error after the cleanup fail:

* stonith-fence_ipmilan-000077880001_start_0 on overcloud-controller-2 'unknown error' (1): call=254, status=Error, exitreason='',
    last-rc-change='Tue Jul 30 08:25:18 2019', queued=0ms, exec=21309ms

This is how Director configurd the fencing:

[heat-admin overcloud-controller-1 ~]$ sudo pcs stonith show stonith-fence_ipmilan-000077880001
 Resource: stonith-fence_ipmilan-000077880001 (class=stonith type=fence_ipmilan)
  Attributes: ipaddr=192.168.104.1 ipport=6233 login=admin passwd=redhat pcmk_host_list=overcloud-controller-0 power_timeout=60
  Operations: monitor interval=60s (stonith-fence_ipmilan-000077880001-monitor-interval-60s)


This is how i can actually fence a node:
ipmitool -I lanplus -U admin -P redhat -H 192.168.104.1 -p 6233 power off

Note that the pcs configuration is lacking the lanplus parameters.

This fix the problem:

sudo pcs stonith update stonith-fence_ipmilan-000077880001 lanplus=true
sudo pcs stonith update stonith-fence_ipmilan-000077880002 lanplus=true
sudo pcs stonith update stonith-fence_ipmilan-000077880003 lanplus=true

Chances are that the way the fence_ipmilan template is used can be improved to automatically add the lanplus parameter.