If your TripleO deployment fails is relatively easy to clean your failed overcloud environment, use heat stack-delete overcloud and Heat will take charge of deleting all the stack, the associated deployments and power off the Ironic nodes.
However there are sometimes when it does not work that way and you will need to manually reset your Ironic nodes. After enduring some pain during a Red Hat OSP Director deployment I decided to document the cleaning process and publish it here.
Power off all Ironic nodes.
Set provision state to available, this was for me the tricky one and took me some trial error tests until I figured out because the parameter for ironic node-set-provision-state is not available but provide, actually there is no available parameter for this Ironic command.
Disassociate the nova instances from the nodes.
Get the nodes out of maintenance state.
List the nodes to verify the executed steps.
Finally clean all the nova instances with nova delete.
Hopes this helps if you are into OpenStack deployments using TripleO. Comments are welcome.