3

I did command om master node

microk8s add-node
From the node you wish to join to this cluster, run the following:
microk8s join 192.168.0.xxx:25000/token

If the node you are adding is not reachable through the default interface you can use one of the following:
 microk8s join 192.168.0.yyy:25000/token
 microk8s join 192.168.0.xxx:25000/token
 microk8s join 172.17.0.1:25000/token

and on another machine i did

microk8s join 192.168.0.xxx:25000/token
Contacting cluster at 192.168.0.xxx
Waiting for this node to finish joining the cluster. .. .. .. .. .. .. .. .. .. ..

it ended without messages but on master node command output microk8s.kubectl get no doesn'show this node only default node

Sultan Zhumatayev
  • 535
  • 1
  • 9
  • 18
  • Please execute command microk8s status and microk8s inspect for a deeper inspection. Did you turn on firewall (ufw)? If yes, consider allowing all node ips like: ufw allow from [xxx.xxx.xxx.xxx](https://xxx.xxx.xxx.xxx/) (repeat for all nodes including the current node)? Did your instances instances having both an internal and public network ? Sometimes they would try talking to the wrong ones. When you will take away the external interfaces, it will work properly. – Malgorzata Mar 04 '21 at 08:44
  • For now i downgrade microk8s to 1.18 and there node joined and exists in node list. – Sultan Zhumatayev Mar 04 '21 at 08:58

2 Answers2

4

Solution may be as @Sultan Zhumatayev said downgrading microk8s to 1.18 and node will join and appear in node list.

However firstly to check where problem lies you can execute commands: $ microk8s status and $ microk8s inspect for a deeper inspection

Another possible solutions:

1. If you turn on firewall (ufw), consider allowing all node ips like: ufw allow from xxx.xxx.xxx.xxx (repeat for all nodes including the current node). If your instances instances having both an internal and public network, they could try talking to the wrong ones. After taking away the external interfaces, it will work properly. Take a look: microk8s-join-stuck.

2. Try to disconnect the node from the cluster. Run on the node the following command:

$ sudo microk8s leave

Then you can try to connect it again, by running add-node on the master:

$ sudo microk8s add-node

and then the respective command on the node again:

$ sudo microk8s join

Read more: microk8s-join.

See also: microk8s.

Malgorzata
  • 6,409
  • 1
  • 10
  • 27
  • No because I did inspection and there was one warning about iptables then i did fix it,but node didn't join,I tried leave and it was endless did add node many times with different urls had no effect .Only downgrade helped. – Sultan Zhumatayev Mar 05 '21 at 12:49
  • I have mentioned your solution as possible solution but for such problems points 1/2 can also help. I have also given hints for community how to debug and find the cause of problem (proper commands). – Malgorzata Mar 12 '21 at 12:31
  • 1
    Searched all over GitHub and Stackoverflow, this is the best answer. Specifically running `sudo ufw allow from xxx.xxx.xxx.xxx` – Farzad Soltani Jan 08 '22 at 11:10
1

I was also having some trouble listing the microk8s nodes and what worked for me was giving an unique hostname of each of the pies in the /etc/hostname file (the default hostname in all the pies after installation was ubuntu). After I rebooted each of the nodes, they showed up with the microk8s kubectl get no on the master node.

It seems that microk8s uses the hostname as an identifier for a the nodes, or at list the kubectl get node, instead of their ip, which in my case, I would prefer. At least in my case I'm using statics ips in my local network.

Fernando Mata
  • 11
  • 1
  • Can confirm, i have a cluster of 3 Raspberry Pi 4 Bs. With a fresh Ubuntu install they all used the hostname `ubuntu`, so I used `sudo hostnamectl set-hostname ` on every Pi and gave them unique hostnames instead. Now all nodes show up in `kubectl get nodes`. – Martin Jan 20 '22 at 14:43