How do I specify an IP address of a peer (node) for the node-to-node mesh?

Open Source Calico Help
1

My worker nodes have two addresses assigned, one for iSCSI traffic and one for all other traffic. When I installed calico the first worker node peered with the address I expected, but the second is trying to peer with the iSCSI address and is not establishing (as expected).

Standard Network: 10.1.30.0/24
iSCSI network: 10.1.20.0/24

I need to modify 10.1.20.62 to 10.1.30.62. How would I accomplish this?

arozar@k8s-cp-1:~$ sudo ./calicoctl node status
Calico process is running.

IPv4 BGP status
+--------------+-------------------+-------+----------+-------------+
| PEER ADDRESS |     PEER TYPE     | STATE |  SINCE   |    INFO     |
+--------------+-------------------+-------+----------+-------------+
| 10.1.30.55   | node-to-node mesh | up    | 01:34:57 | Established |
| 10.1.30.56   | node-to-node mesh | up    | 01:37:13 | Established |
| 10.1.20.62   | node-to-node mesh | start | 01:37:53 | Connect     |
| 10.1.30.61   | node-to-node mesh | up    | 01:46:34 | Established |
+--------------+-------------------+-------+----------+-------------+
2

So it looks like if I disable this iSCSI address, reboot it uses the proper address. Then use netplan to re-apply the address. But this seems like a possible issue in the event of a reboot. Is it possible to bind node-to-node mesh to a particular address?

UPDATE -
As soon as I re-apply with netplan it roles back over to the iSCSI address.

3

If you have multiple interfaces, you’ll need to set the IP autodetection method to pick up the right one. See this doc: Configure IP autodetection

Without that, it picks up the first IP that it finds.

4

Thank you @fasaxc, works like a charm.

The solution:
kubectl set env daemonset/calico-node -n kube-system IP_AUTODETECTION_METHOD=can-reach=10.1.30.1