tailscale: K8s Operator: Cross cluster Ingress / Egress not working
What is the issue?
I am trying to get expose a service in cluster A to cluster B.
Cluster A has a service setup like so.
apiVersion: v1
kind: Service
metadata:
name: argocd-server
annotations:
tailscale.com/expose: "true"
tailscale.com/hostname: argocd-dev
spec:
type: ClusterIP
ports:
- port: 80
targetPort: 8080
protocol: TCP
name: http
- port: 443
targetPort: 8080
protocol: TCP
name: https
I am able to hit this service via it’s tailscale magic dns name just fine from local while on tailscale.
Cluster B has a service set up like…
apiVersion: v1
kind: Service
metadata:
annotations:
tailscale.com/tailnet-fqdn: argocd-dev.TAILNET_NAME.ts.net
name: argocd-dev
spec:
externalName: placeholder
type: ExternalName
From looking at the operator logs as well as the egress proxy logs, everything seems to be relatively happy.
However, when I try to just ping the new Tailscale service that’s spun up, the traffic is not being routed to Cluster A and all packets get dropped.
Steps to reproduce
No response
Are there any recent changes that introduced the issue?
No response
OS
Linux
OS version
No response
Tailscale version
1.62.0
Other software
This is running inside a newer GKE cluster. I am NOT using Autopilot.
Bug report
No response
About this issue
- Original URL
- State: open
- Created 3 years ago
- Comments: 67 (27 by maintainers)
Def not using Autopilot clusters. Those are garbage. Haha. In your GKE cluster page, under the networking tab… What do you have for DNS provider?
Given that I have a work around for now, I think I’ll just wait. Im still a pretty big networking noob, and it sounds like I would more likely mess it up and cause more noise for you. But once there is a release available, I’ll get noisy again (;
I’ll keep an eye out on that thread and as soon as I catch wind that its available in some type of release I’ll be more than happy to give it a shot.
I’ll grab the latest version off of
https://pkgs.tailscale.com/unstable/helmcharts@irbekrm It doesn’t look like 1.63.1 has been uploaded to
https://pkgs.tailscale.com/helmchartsfortailscale-operatorThis is run from within cluster B. And this makes sense to me because cluster B doesn’t exist “within” the tailnet.
argocd-dev-tsis the name of the service in Cluster B with theExternalNameattached to it so it can reach cluster A.