How the FelixConfiguration is being recreated and how to change its default values? #9147
-
Hello Team, I would like to understand the behavior of recreating FelixConfiguration resource. I have reinstalled calico/tigera to version v3.28.1 on my k8s cluster. By default it created following FelixConfiguration resource:
This configuration results in following log entry:
I have modified the spec entry locally and reapplied it via calicoctl to following values:
This was accepted and after restart of calico-nodes the log was changed to:
When I have tried to describe the FelixConfiguration once again, I found that both options were present, 'routeTableRanges' and my previous 'routeTableRange' which I tried to remove. Then I tried to remove the FelixConfiguration manually and it was immediately recreated with its default values which I posted at the top of this post. What is recommendation for configuring this part to be upgrade-safe when the previous field will be removed? How to get rid of the warning in the log? |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment
-
FelixConfiguration will have a default resource created by calico-node and/or operator. I've created tigera/operator#3757 for this issue. |
Beta Was this translation helpful? Give feedback.
FelixConfiguration will have a default resource created by calico-node and/or operator.
If you are using AmazonVPC CNI plugins then the operator will set a default RouteTableRange, so I would guess it is the operator that keeps setting that field. (the operator also sets it if using GKE but due to the values you're seeing I'm assuming you're using AmazonVPC)
You should be able to see the CNI type in the Installation resource (
kubectl get installation default -o yaml
).I've created tigera/operator#3757 for this issue.