Flux Pods not working for Liquid Metal / capmvm Leaf cluster - wrong IPs

[LutzLange]

What happened: [A clear and concise description of what the bug is.] The cilium deployment did not restart my flux pods. Thus they can't reach coredns and my profiles will not deploy correctly. Restarting the flux pods solves the issue.

This is the situation after my initial deployment :

I'm restarting the flux pods : kubectl --kubeconfig=/home/lutz/Downloads/lutzmvm16.kubeconfig get pods -A -o wide | grep 88 | awk '{ print $2 }' | while read POD; do kubectl --kubeconfig=/home/lutz/Downloads/lutzmvm16.kubeconfig delete pod $POD -n wego-system; done

This will replace the 10.88 addresses and bring flux back into a working stage.

And a little while later, my prometheus profile is deployed as well :

What did you expect to happen: I need leaf cluster bootstrapping to work and flux to come up in working condition.

I think that I'm facing a timing issue here. The flux bootstrapping as applied before cilium comes up successfully. How could we make the bootstrapping depend on successful CNI deployement?

How to reproduce it: Contact me. I can and will give you access to my demo environment. https://demo.weavegitops.com/

I'm using a management cluster on EKS. And Liquid metal with Equinix Metal service.

This is my management repository : https://github.com/weavegitops/demo-repo I can invite you to the weavegitops org, please ping me on Slack.

Environment:

• capmvm version: 0.3.0
• WGE: 0.18 aka 0.0.17-56-gffbf05f