Kubernetes services not starting up after reboot

[kodaren]

Hi,

I installed on a clean Ubuntu Server 18.04 on bare metal. It worked fine until I rebooted. The KUBECONFIG variable is set properly. Then I get the dreadful:
The connection to the server 127.0.0.1:6443 was refused - did you specify the right host or port?

I reinstalled. First cleanup.sh, removed .wks folder and then ran setup.sh again. Working fine, but does not survive a reboot! Same as before

Why?

[chanwit]

If you're using Ignite backend, it's a known bug in Ignite.

[stealthybox]

Thanks for playing with this @kodaren.
It's very reasonable to assume that since you installed Firekube on a node, that when you would reboot it, kubernetes and all of the ignite vm's would still be running.

Some design choices that we are working through currently mean that this does not happen.
We discussed this issue in depth on the ignite dev call: nov4 ignite dev notes

in short

• ignite does not support restart-policy: always a la docker
• ignite's ignited implementation is not finished
• ignite vm's IP addr's are not theoretically stable. (in practice, they may be quite stable depending on usage)

---

possible workaround
Could you please cd to your firekube repo and run footloose start after a reboot and see if the vm's come up and that they have the same IP addresses?
( If so, you could create a systemd unit that does this :P )

Would you have the time to join next week's November 11 ignite dev call to talk about how you think this should all work?

Thanks so much :)

[kodaren]

Hi,

It is still not working.

1. The binaries for footloose and ignite are in the ~/.wks/bin folder. However, If you add that folder to the path then you still cannot run footloose as it requires sudo and it won't find the binaries in that path by default. I copied the files to the /usr/local/bin folder and then they are accesible

2. I ran footloose start in the git repo and two vm:s were created

3. footloose show display the two nodes. docker ps show two ignite containers

4. However, I get no response when running any kubectl command like get nodes or get po. It just waits forever. (Not connection refused). Something wrong with the network?

5. If I run sudo kubectl get po then I get:
   The connection to the server localhost:8080 was refused - did you specify the right host or port?

[stealthybox]

@kodaren, you would need to use the proper KUBECONFIG to talk to the cluster.

Here's the default wks-quickstart location:

export KUBECONFIG=$HOME/.wks/weavek8sops/example/kubeconfig

You set this before you started using your cluster.