This message was deleted.
# k3s
a
This message was deleted.
👀 1
b
Can you explain what happened first? Was there a series of steps that arrived at the ConfigMap no longer working?
m
we had an upgrade of the underline servers, and I guess k3s service has been restarted and went back to it's default conf of coredns
When I checked the coredns configmap, I couldn't see anymore the nameserver I added before the upgrade
b
Um, this is an interesting problem. k3s ships with coredns manifests instead of a helmchart. That means that if you don't prep the OS with your own
/var/lib/rancher/k3s/server/manifests/coredns.yaml
file, your configmap will be over written when starting a new node.
👍 1
it might even happen with a k3s restart
and this goes for every node
You could maybe make this easier on people in the future by appealing to k3s to embed the chart like it does for traefik and somehow using that, which would have more predictable behavior with no conflicting
HelmChartConfig
resources. But I don't know how open they would be to the idea - I assume they are doing the current approach because of some bootstrapping chicken-and-egg issue
The short term solution is to either: • modify that file before k3s starts • delete the entire directory (i haven't verified this, but i think it should work fine)
m
How about deploying a custom configmap for coredns and have coredns merge it to the main conf shiped by k3s ? I rmemeber that I came across this feature one day but I don't remeber how it can be configured
b
yeah i haven't heard about any such function. sounds interesting. if you figure it out, let me know 🙂
m
sure will do