workers on all nodes

[maxcerny]

Simple question. I have created 5 node clusters that have 3 master 2 worker nodes no problems, but is there a way to create a cluster that has workers on all nodes? Is it as easy as adding the nodes in both master and worker?

[timothystewart6]

This should be as simple as removing this line right here in your all.yaml (line 36)

  {{ '--node-taint node-role.kubernetes.io/master=true:NoSchedule' if k3s_master_taint else '' }}

This applies a taint to a node so that it is marked as a master and no workload will run unless they are critical. If you've already created your cluster, you can remove that taint

kubectl taint nodes node1 node2 node3  node-role.kubernetes.io/master=true:NoSchedule-

(You'll need to updated the node names to match yours)

if you notice the - at the end, that means remove, so if you need to add the taint back you can run the same command without the -

[maxcerny]

Amazing, thanks

[mlanner]

I have a similar issue. I've got three nodes and I've put all of them in both the [master] and [node] sections, like so:

[master]
node01 ansible_ssh_host=10.10.1.51 
node02 ansible_ssh_host=10.10.1.52 
node03 ansible_ssh_host=10.10.1.53

[node]
node01 ansible_ssh_host=10.10.1.51 
node02 ansible_ssh_host=10.10.1.52 
node03 ansible_ssh_host=10.10.1.53 

[k3s_cluster:children]
master
node

I originally didn't catch the {{ '--node-taint node-role.kubernetes.io/master=true:NoSchedule' if k3s_master_taint else '' }} part and ran the playbook with it, tainting the nodes. This resulted in a broken installation where, towards the end, I get fatal errors:

TASK [k3s_agent : Enable and check K3s service]
fatal: [node02]: FAILED! => {"changed": false, "msg": "Unable to start service k3s-node: Job for k3s-node.service failed because the control process exited with error code.\nSee \"systemctl status k3s-node.service\" and \"journalctl -xeu k3s-node.service\" for details.\n"}

Checking on the nodes and trying to restart the k3s-node.service just keeps on failing. And I'm not seeing anything useful in the systemctl or journalctl output.

I then found this thread and ran the reset.yml playbook. After that I removed the above from extra_server_args and re-deployed. However, I'm still having the same issue with the k3s node service not starting.

On line 26 in all.yml, should I change it to read?

k3s_master_taint: false

Any thoughts on this and where I might have gone wrong?