EDIT: The whole point of my setup is to achieve (if possible) the following :
- I have multiple k8s nodes
- When I contact an IP address (from my company's network), it should be routed to one of my container/pod/service/whatever.
- I should be able to easily setup that IP (like in my service .yml definition)
I'm running a small Kubernetes cluster (built with kubeadm) in order to evaluate if I can move my Docker (old)Swarm setup to k8s. The feature I absolutely need is the ability to assign IP to containers, like I do with MacVlan.
In my current docker setup, I'm using MacVlan to assign IP addresses from my company's network to some containers so I can reach directly (without reverse-proxy) like if it's any physical server. I'm trying to achieve something similar with k8s.
I found out that:
- I have to use Service
- I can't use the LoadBalancer type, as it's only for compatible cloud providers (like GCE or AWS).
- I should use ExternalIPs
- Ingress Resources are some kind of reverse proxy ?
My yaml file is :
apiVersion: apps/v1beta1
kind: Deployment
metadata:
name: nginx-deployment
spec:
template:
metadata:
labels:
app: nginx
spec:
containers:
- name: nginx
image: nginx:1.7.9
ports:
- containerPort: 80
nodeSelector:
kubernetes.io/hostname: k8s-slave-3
---
kind: Service
apiVersion: v1
metadata:
name: nginx-service
spec:
type: ClusterIP
selector:
app: nginx
ports:
- name: http
protocol: TCP
port: 80
targetPort: 80
externalIPs:
- A.B.C.D
I was hopping that my service would get the IP A.B.C.D (which is one of my company's network). My deployment is working as I can reach my nginx container from inside the k8s cluster using it's ClusterIP.
What am I missing ? Or at least, where can I find informations on my network traffic in order to see if packets are coming ?
EDIT :
$ kubectl get svc
NAME CLUSTER-IP EXTERNAL-IP PORT(S) AGE
kubernetes 10.96.0.1 <none> 443/TCP 6d
nginx-service 10.102.64.83 A.B.C.D 80/TCP 23h
Thanks.
you can try kube-keepalived-vip configurtion to route the traffic. https://github.com/kubernetes/contrib/tree/master/keepalived-vip
A solution that could work (and not only for testing, though it has its shortcomings) is to set your Pod to map the host network with the
hostNetwork
spec field set totrue
.It means that you won't need a service to expose your Pod, as it will always be accessible on your host via a single port (the
containerPort
you specified in the manifest). No need to keep a DNS mapping record in that case.This also means that you can only run a single instance of this Pod on a given node (talking about shortcomings...). As such, it makes it a good candidate for a DaemonSet object.
If your Pod still needs to access/resolve internal Kubernetes hostnames, you need to set the
dnsPolicy
spec field set toClusterFirstWithNoHostNet
. This setting will enable your pod to access the K8S DNS service.Example:
EDIT: I was put on this track thanks to the the ingress-nginx documentation