Andrii Miroshnychenko

I have vanilla Kubernetes 1.26.1, Calico v3.25.0, MetalLB 0.13.7 Kernel Ubuntu 22.04.1 5.15.0-58-generic. Cluster has 3 master nodes and 3 worker nodes. My ISP delegated `a.b.c.184/29` subnet to me where...

What I see with tcpdump at `enp6s19`: ``` 02:04:54.858291 IP x.y.z.230.55215 > a.b.c.186.443: Flags [S], seq 1693191197, win 65535, options [mss 1460,nop,wscale 6,nop,nop,TS val 266762460 ecr 0,sackOK,eol], length 0 02:04:54.858847...

In my case the issue was in source based routing. I had the rule `iptables -t mangle -A PREROUTING -i enp6s19 -j MARK --set-mark 1019` and this made the second...

@johnsimcall yes, I did. I had a mark rule with iptables and an `ip rule` which caused the problem. Because of this the second packet (VXLAN traffic) had been sent...

@johnsimcall do you want the configs?

@johnsimcall As I mentioned before - I have 6 nodes in the cluster. API is available via dedicated IP (it could be haproxy, kube-vip in my case). Cluster was installed...

@johnsimcall Source Based Routing. Worker nodes does not handle the dedicated routable IPv4 number. `enp6s19` is just UP without any IP at all: ``` # cat /etc/network/interfaces auto enp6s19 iface...

No plans to add that into the docs?

Hi @srikumar003. Had you a chance to check the PR?

I have a couple of load balancers and would like to use that feature