Ladies and gents, was hoping to get some clarification on the setup of consul agent running on a Windows VM, with regards to multiple NIC’s and an F5 integration.
The way the VM is set up, is:
Two network interfaces, the first one is only routable from the F5 load balancer and has no gateway out. When the traffic reaches the VM from the F5, it goes out via the second interface, which has a default gateway and is generally a routable VM address.
F5 populates pool members from the https://www.consul.io/api-docs/catalog#list-nodes-for-service method, and I’m trying to figure out how to advertise the first “unroutable” network interface to consul catalog, while keeping all communication via the second “normal” interface.
After some research, I ended up with a config file (some non-relevant entries omitted):
{
“server”: false,
“datacenter”: “tcb”,
“node_name”: “tcb-winweb261”,
“bind_addr”: “172.20.14.161”,
“serf_lan”: “172.20.14.161”,
“serf_wan”: “172.20.14.161”,
“advertise_addr”: “172.20.13.161”,
“client_addr”: “127.0.0.1”,
“retry_join”: [“172.20.12.41”, “172.20.12.42”, “172.20.12.43”]
}
Here, the 172.20.13.0/24 is the first and “unroutable” network which I actually need in the catalog for the F5 to pick it up, and the 172.20.14.0/24 is the normal “routable” network, which I need to use for cluster communication.
With the config above applied, the Serf health seems to be stuck in a flapping state, as per documentation. However, shouldn’t serf_lan be taken into account by the Consul servers? Is it ignored somehow?
Is such a working setup at all possible with Consul? The log output does not seem to produce anything meaningful.
Would very much appreciate any input here!