When the openshift-install agent wait-for bootstrap-complete command cannot connect to either the k8s API or the assisted-service API, it tries to ssh to the rendezvous host to see if it is up.

If there is a running ssh-agent on the local host, we connect to it to make use of its private keys. This is not guaranteed to work, as the private key corresponding to the public key in the agent ISO may not be present on the box.

If there is no running ssh-agent, we use the literal public key as the path to a file that we expect to contain the private key. This is guaranteed not to work.

All of this generates a lot of error messages at DEBUG level that are confusing to users.

If we did succeed in ssh-ing to the host when it has already joined the cluster, the node would end up tainted as a result, which we want to avoid. (This is unlikely in practice though, because by the time the rendezvous host joins, the k8s API should be up so we wouldn't normally run this code at that time.)

We should stop doing all of this, and maybe just ping the rendezvous host to see if it is up.