https://rancher.com/ logo
Join Slack
Powered by
This message was deleted.
# k3s
a
This message was deleted.
a
Add additional node pools to reflect your new architecture, drain & remove old nodes.
c
you mean i have to create new control nodes and remove the old ones?
wanted to avoid that as usually i break etcd when doing that 😉
b
@curved-army-69172 just taint the nodes?
c
The worker role means nothing. Its presence or absence does not have any influence on whether or not workloads are allowed to run there. If you don't want workloads on those nodes, you should taint them.
👍 1
c
oh, may I ask why there is a role at all then? When I taint with no schedule - do all control plane related deployments (including ingress and such) have the appropriate tolerations? sorry if asking dumb questions here, just getting my feet wet 😉
c
historical reasons, mostly.
some people like to see a role on their agent nodes
some people use the label as a selector for workload scheduling purposes
in reality you can make up whatever role labels you want
and then use those in nodeselectors on your workload pods
but its up to you to use those how you want
if you want to actually keep things off the servers, you need to taint them
you can add that taint after the fact with 
kubectl taint node
c
thanks, yeah did that now - would need to go over all the "default" deployments on the control-plane to see if all have the proper tolerations
and figure out how to add the taint to the whole node pool 😉 so new nodes get it automatically just in case we need to replace the nodes
Open in Slack
PreviousNext
Powered by