Only 1 worker is launched in a multi-node vagrant setup

[F21]

I am using the multi-node vagrant setup and have configured it to create 3 worker nodes. I can see that the nodes are launched and can ssh into all of them. However, kubernetes only sees 1 worker:

$ kubectl get nodes
NAME           STATUS                     AGE
172.17.4.101   Ready,SchedulingDisabled   7m  <--- controller
172.17.4.201   Ready                      7m               <--- worker

When I ssh into worker 2 and 3 (the ones that were not running kubernetes), I see that there are not logs for the kubelet:

core@w3 ~ $ journalctl -u kubelet
-- No entries --

This is also the case when I check the web dashboard.

I am using core-kuberntes master.

This is my config.rb:

$update_channel="alpha"

$controller_count=1
$controller_vm_memory=512

$worker_count=3
$worker_vm_memory=2048

$etcd_count=3
$etcd_vm_memory=512

I also set USE_CALICO=true and K8S_VER=v1.4.5_coreos.0 in worker-install.sh and controller-install.sh.

I am running this in VirtualBox 5.1.8 on Windows 10 64-bit. To get vagrant up working on Windows, I applied this commit to master: https://github.com/ah45/coreos-kubernetes/commit/3ca05ebbf21610401781dc2410293636e20d6161

[F21]

It seems to be pretty random whether a worker will be initialized correctly or not. I just destroyed the cluster and ran vagrant up again. This time, I can see 2 workers downloading the images and setting up, but the last worker did nothing. Is there anyway to check to see why the last worker did not install?

Using journalctl just showed no entries:

core@w3 ~ $ journalctl -u kubelet
-- No entries --