debug cozy-proxy

[RaSerge]

Hello, cozydevs!
I deploy cozy-proxy v0.1.4 with cozystack on my kubeadm based clusters. Every HelmRelease is ready. Metallb works and can assign IPs for Services.

I create tenant without "isolated", then create vm-disk and then vm-instance.

when i deploy vm-instance as external with externalMethod: PortList vm have acces to external http server (hostname resolved, ping is ok, curl is ok)
When i deploy vm-instance as external with externalMethod: WholeIP vm don't have acces to external http server (hostname resolved, but curl connection is stuck. no ping. When i try to do traceroute from guest OS i see that traceroute stops at first hop - kubeovn interface)

How to debug this connectivity problem? or it is better to update to latest cozy-proxy v0.2.0? (but is no any bugfixes in changelog)

[kvaps]

Hey,

I deploy cozy-proxy v0.1.4 with cozystack on my kubeadm based clusters. Every HelmRelease is ready. Metallb works and can assign IPs for Services.

You don't need to deploy it manually, cozy-proxy already running in cozy-system namespace

Please check that you have no conflict of two running cozy-proxy daemonsets.

How to debug this connectivity problem?

To debug you can use kubectl node-shell plugin to access host namespace of your node:

kubectl node-shell <node> -x -n kube-system
apk add tcpdump
tcpdump -i any host <vm_ip> or host <vm_external_ip>

Also check:

apk add nftables
nft list table ip cozy_proxy

on the node where you run the vm.

Please check MTU on the interface inside the VM, it should be 1400

is better to update to latest cozy-proxy v0.2.0

cozy-proxy introduces new method for mapping IP address, not it uses performs nat via raw mangle table with allows to make router decision after rewriting IP