It's actually not so easy.

The fastest way to recover  is just restore  from backup.
Otherwise  the flow  should  be:
1.  Install the new  node  (new hostname will be  easier).
2.  Use gluster's  replace-brick to change the dead brick with new one.
3.  Once ovirt 's  integration with gluster  detects the change, you will be able to forcefully remove  the  dead node.
4. Add  the newly installed node  to the relevant cluster (with or  without hosted-engine deployment)
5.  Test to move a low-priority VM to the new  host.
6.  Power  up  a test VM on the new host to test the functionality.
7. You can make the new node SPM and test snapahots and new vm creation.

As  you can see , in order to remove a missing node  - it should not be part of the Gluster Cluster.

Best Regards,
Strahil Nikolov