I am trying to connect to an overlay kube network, but I'm getting very strange results. Trying to follow along their docs https://docs.cilium.io/en/stable/network/vtep/
the equivalent linux configuration would be:
Code: Select all
Test VTEP Integration
Node IP: 10.169.72.233
+--------------------------+ VM IP: 10.169.72.236
| | +------------------+
| CiliumNode | | Linux VM |
| | | |
| +---------+ | | |
| | busybox | | | |
| | | ens192<------>ens192 |
| +--eth0---+ | | |
| | | +-----vxlan2-------+
| | |
| lxcxxx |
| | |
+------+-----cilium_vxlan--+
# vxlan interface:
ip link add vxlan2 type vxlan id 2 dstport 8472 local 10.169.72.236 dev ens192
ip link set dev vxlan2 address 16:84:C6:DB:31:57
ip link set vxlan2 up
# Configure the VTEP with IP 10.1.1.236 to handle CIDR 10.1.1.0/24.
ip addr add 10.1.1.236/24 dev vxlan2
# Assume Cilium podCIDR network is 10.0.0.0/16, add route to 10.0.0.0/16
ip route add 10.0.0.0/16 dev vxlan2 proto kernel scope link src 10.1.1.236
# Allow Linux VM to send ARP broadcast request to Cilium node for busybox pod
# ARP resolution through vxlan2 device
bridge fdb append 00:00:00:00:00:00 dst 10.169.72.233 dev vxlan2
1 vxlan interface with source address 10.169.72.236 and vni 2 and mac:16:84:C6:DB:31:57
1 Ip address 10.1.1.236/24 to the vxlan interface
1 vtep for each node in the cluster
The part where I think I'm getting stuck is to do the equivalent of:
ip route add 10.0.0.0/16 dev vxlan2 proto kernel scope link src 10.1.1.236
bridge fdb append 00:00:00:00:00:00 dst 10.169.72.233 dev vxlan2
I tried adding an IP address of the range 10.0.0.0/16 to the interface, for few seconds works, then starts getting dupped packages and stops working.
Any idea?