SKALA NETWORKS

@zbb88888 So you're suggesting having a Subnet dedicated to only having VIPs inside right? It could work to have IPs withins the VPC assigned to LoadBalancers. And NAT gateways in...

I'm still puzzled by what causes this behaviour, using a dedicated Subnet is a bit more complex than using an already existing one. For my use cat of doing NAT...

@oilbeater Should we document `switch_lb_vip` to signal they might be broken? It seems evident the implementation is correct in Kube-OVN, but OVN does something behind the scenes with the logical...

Considering the VIP system seems to have a bug, I tried using a separate subnet. Basically, I have a VM in subnet A and a subnet B that has nothing...

Removing the subnet but keeping the route on the NAT GW to send the traffic directed to the LB on the default gateway yields the same problem. The trace isn't...

Address conflict is bypassed by live migration flags in the code. So I guess it is totally doable to have 2 IP Custom Resources pointing to different pods.

It could be doable to auto-patch the content of the CRD "VirtualMachineSnapshotContent", exporting all the kubeovn.io annotations from the virt-launcher pod back into it. Considering these annotations on a virt-launcher...

@morete Just a few questions considering I can't even make it work with Pods (nor VMs) in v1.14.0: - Do you define the VIP before hand (as in create a...

Got it, I'm interested in this bug because I have the same usecase (LB on VMs). I indeed found a bug with the IPFamily of the services, but it must...

I can reproduce the bug. @morete the bot is hallucinating and confusing ServiceMonitor with Prometheus CRDs. I don't think it's on the right track at all.