You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository has been archived by the owner on Dec 6, 2024. It is now read-only.
I setup linkerd-tcp to run on my cluster alongside Linkerd in a CNI configuration as a Daemonset. In my Kubernetes version 1.5.1, using hostNetwork: true doesn't give access to Kubernete's DNS options(for example, namerd.default.svc.cluster.local) A solution to this would be to expose the web port from Namerd as a NodePort service, however, this would require using the downward API env variable "NODE_NAME" to be able to access the Namerd NodePort service. The problem is that the downward API isn't available for ConfigMap resources.
A workaround was to hardcode a LoadBalancer service URL as the namerd url config for Linkerd-TCP.
I think that a configuration option to allow the url to be built from "NODE_NAME" env variable would solve this more elegantly as a solution to running Linkerd-TCP as a DaemonSet with CNI. Maybe a "fromNodePort" or something.
Of course, this might not be needed with Kubernetes 1.6 because they added a "ClusterFirstWithHostNet" dnsPolicy option which keeps DNS settings even with hostNetwork: true. I haven't updated yet so I haven't played with that option.
The text was updated successfully, but these errors were encountered:
@FuzzOli87 do you have an example of the k8s artifact or configuration on how you got linkerd-tcp setup alongside linkerd in a CNI config DaemonSet? I can't seem to find any examples. Would greatly appreciate it.
Sign up for freeto subscribe to this conversation on GitHub.
Already have an account?
Sign in.
I setup linkerd-tcp to run on my cluster alongside Linkerd in a CNI configuration as a Daemonset. In my Kubernetes version 1.5.1, using
hostNetwork: true
doesn't give access to Kubernete's DNS options(for example, namerd.default.svc.cluster.local) A solution to this would be to expose the web port from Namerd as a NodePort service, however, this would require using the downward API env variable "NODE_NAME" to be able to access the Namerd NodePort service. The problem is that the downward API isn't available for ConfigMap resources.A workaround was to hardcode a LoadBalancer service URL as the namerd url config for Linkerd-TCP.
I think that a configuration option to allow the url to be built from "NODE_NAME" env variable would solve this more elegantly as a solution to running Linkerd-TCP as a DaemonSet with CNI. Maybe a "fromNodePort" or something.
Of course, this might not be needed with Kubernetes 1.6 because they added a "ClusterFirstWithHostNet" dnsPolicy option which keeps DNS settings even with hostNetwork: true. I haven't updated yet so I haven't played with that option.
The text was updated successfully, but these errors were encountered: