chrisarnott86

@Cellebyte, @thebsdbox I've done more testing on a new test cluster to try and eliminate as many variables as possible. Using a barebones debian (not configured with our org's standard...

@Cellebyte and @thebsdbox, I recently installed version 0.9.1 in our production cluster, I had to remove it after I was seeing similar connection dropping issues to previously with earlier versions....

Thanks @Cellebyte, sorry for the slow reply. I've done a bit more testing and I've found what I think may be the issue (although not what the cause is). I...

Hi @Cellebyte and @thebsdbox. After a long break from looking at this (also running the system in a dev cluster for testing). I implemented this in my prod cluster again....

Thanks @Cellebyte, networking is the default k3s setup, so that's flannel as cni I believe? Here's the service yaml for the testing egress. ``` apiVersion: v1 items: - apiVersion: v1...

Thaks @Cellebyte. The cluster setup is 3 control plane/master nodes and 1 worker. So 1 kube-vip pod on each master plus the 1 additional pod for the cloud controller. I'm...

In my testing example, I've got the pod running on the same node as the elected leader for the service (as far as I can see) but there are still...

I've done a further test whereby I reduced my dev cluster to one node, I still got the same packet loss issues.

Some more testing has revealed that the packets are being lost coming back into the egress from the remote end. So if I ping out from the pod behind the...

OK, I've got to the bottom of it, I had accidentally used the same IP address in both my dev and production clusters for the egresstest workload. So it's all...