Submariner is a tool built to connect overlay networks of different Kubernetes clusters. While most testing is performed against Kubernetes clusters that have enabled Flannel/Canal, Submariner should be compatible with any CNI-compatible cluster network provider, as it utilizes off-the-shelf components such as strongSwan/Charon to establish IPsec tunnels between each Kubernetes cluster.
Note that Submariner is in the pre-alpha stage, and should not be used for production purposes. While we welcome usage/experimentation with it, it is quite possible that you could run into severe bugs with it, and as such this is why it has this labeled status.
The basic architecture diagram for Submariner is as follows:
Submariner consists of a few components that work and operate off of Kubernetes Custom Resource Definitions (CRDs). The Submariner CRDs are clusters.submariner.io
and endpoints.submariner.io
.
The two primary Submariner components within connected clusters are:
- submariner (Deployment)
- submariner-route-agent (DaemonSet)
The submariner pods are run on the gateway nodes, and will perform leader election between them to elect an active IPsec endpoint. On gateway nodes that are idle, the submariner-route-agent
will configure routes to enable full connectivity to remote clusters from non-leader gateway and worker nodes. submariner-route-agent
as a component is run on every node, and is aware of the current leader. When run on the gateway node that is the leader, it will simply sit idle awaiting leader loss.
Upon startup, the submariner
pod that is elected leader will perform a reconciliation process that ensures it is the sole endpoint for this cluster. This is part of the reason why it is important to have unique cluster IDs between clusters, as two clusters with the same ID will reconcile each other out of existence.
Upon failure, another Submariner pod (on one of the other gateway hosts) will gain leadership and perform reconciliation to ensure it is the active leader. When done, the remote clusters will reconcile the IPsec endpoint to the new endpoint, and connection will be re-established. In the interim, the submariner-route-agent
pods will update the route tables on each node to point towards the new endpoint host.
Submariner uses a central broker to facilitate the exchange of information and sync CRD's between clusters. The datastoresyncer
runs as a controller within the leader-elected submariner
pod, and is responsible for performing a two-way synchronization between the datastore and local cluster of Submariner CRDs. The datastoresyncer
will only push CRD data to the central broker for the local cluster (based on cluster ID), and will sync all data from the broker the local cluster when the data does not match the local cluster (to prevent circular loops)
submariner (compiled as the binary submariner-engine
) has a few controllers built into it that establish state. It is responsible for running/interfacing with Charon to establish IPsec tunnels, as well as updating local cluster information into the central broker to share information between clusters.
submariner-engine runs and utilizes leader election to establish an active gateway node, which is used to facilitate IPsec tunnel connections to remote clusters. It also manipulates the IPtables rules on each node to enable a. forwarding of traffic and b. SNAT for local node traffic.
The submariner-route-agent runs as a DaemonSet on all Kubernetes nodes, and ensures route rules to allow all pods/nodes to communicate through the elected gateway node for remote cluster networks. It will ensure state and react on CRD changes, which means that it is able to remove/add routes as leader election occurs.
The network path of Submariner varies depending on the origin/destination of the IP traffic. In all cases, traffic between two clusters will transit between the leader elected (in each cluster) gateway nodes, through ip xfrm
rules. Each gateway node has a running Charon daemon which will perform IPsec keying and policy management.
When the source pod is on a worker node that is not the elected gateway node, the traffic destined for the remote cluster will transit through the kernel routing rules table to the cluster-local gateway node, which will perform source network address translation (SNAT) to the remote network. This allows for much more efficient traffic selectors to be configured, as well as more predictable routing paths. Once the traffic reaches the destination gateway node, it is routed one of two ways, depending on the destination CIDR. If the destination CIDR is a pod network, the traffic is routed however the CNI-compatible network routes traffic destined for pod IPs. If the destination CIDR is a service network, then traffic is routed through the facility configured via kube-proxy
on the destination gateway node.
Submariner has a few requirements in order to get started:
- At least 3 Kubernetes clusters, one of which is designated to serve as the central broker that is accessible by all of your connected clusters; this can be one of your connected clusters, but comes with the limitation that the cluster is required to be up in order to facilitate interconnectivity/negotiation
- Different cluster/service CIDR's (as well as different kubernetes DNS suffixes) between clusters. This is to prevent traffic selector/policy/routing conflicts.
- Direct IP connectivity between instances through the internet (or on the same network if not running Submariner over the internet). Submariner supports 1:1 NAT setups, but has a few caveats/provider specific configuration instructions in this configuration.
- Knowledge of each cluster's network configuration
- Helm version that supports crd-install hook (v2.12.1+)
An example of three clusters configured to use with Submariner would look like the following:
Cluster Name | Provider | Cluster CIDR | Service CIDR | DNS Suffix |
---|---|---|---|---|
broker | AWS | 10.42.0.0/16 | 10.43.0.0/16 | cluster.local |
west | vSphere | 10.0.0.0/16 | 10.1.0.0/16 | west.local |
east | AWS | 10.98.0.0/16 | 10.99.0.0/16 | east.local |
Submariner utilizes the following tools for installation:
kubectl
helm
base64
cat
tr
fold
head
These instructions assume you have a combined kube config file with at least three contexts that correspond to the respective clusters. Thus, you should be able to perform commands like
kubectl config use-context broker
kubectl config use-context west
kubectl config use-context east
Submariner utilizes Helm as a package management tool.
Before you start, you should add the submariner-latest
chart repository to deploy the Submariner helm charts.
helm repo add submariner-latest https://releases.rancher.com/submariner-charts/latest
The broker is the component that Submariner utilizes to exchange metadata information between clusters for connection information. This should only be installed once on your central broker cluster. Currently, the broker is implemented by utilizing the Kubernetes API, but is modular and will be enhanced in the future to bring support for other interfaces. The broker can be installed by using a helm chart.
First, you should switch into the context for the broker cluster
kubectl config use-context <BROKER_CONTEXT>
If you have not yet initialized Tiller on the cluster, you can do so with the following commands:
kubectl -n kube-system create serviceaccount tiller
kubectl create clusterrolebinding tiller \
--clusterrole=cluster-admin \
--serviceaccount=kube-system:tiller
helm init --service-account tiller
Wait for Tiller to initialize
kubectl -n kube-system rollout status deploy/tiller-deploy
Once tiller is initialized, you can install the Submariner K8s Broker
helm repo update
SUBMARINER_BROKER_NS=submariner-k8s-broker
helm install submariner-latest/submariner-k8s-broker \
--name ${SUBMARINER_BROKER_NS} \
--namespace ${SUBMARINER_BROKER_NS}
Once you install the broker, you can retrieve the Kubernetes API server information (if not known) and service account token for the client by utilizing the following commands:
SUBMARINER_BROKER_URL=$(kubectl -n default get endpoints kubernetes -o jsonpath="{.subsets[0].addresses[0].ip}:{.subsets[0].ports[0].port}")
SUBMARINER_BROKER_CA=$(kubectl -n ${SUBMARINER_BROKER_NS} get secrets -o jsonpath="{.items[?(@.metadata.annotations['kubernetes\.io/service-account\.name']=='${SUBMARINER_BROKER_NS}-client')].data['ca\.crt']}")
SUBMARINER_BROKER_TOKEN=$(kubectl -n ${SUBMARINER_BROKER_NS} get secrets -o jsonpath="{.items[?(@.metadata.annotations['kubernetes\.io/service-account\.name']=='${SUBMARINER_BROKER_NS}-client')].data.token}"|base64 --decode)
These environment variables will be utilized in later steps, so keep the values in a safe place.
Submariner is installed by using a helm chart. Once you populate the environment variables for the token and broker URL, you should be able to install Submariner into your clusters.
-
Generate a Pre-Shared Key for Submariner. This key will be used for all of your clusters, so keep it somewhere safe.
SUBMARINER_PSK=$(cat /dev/urandom | LC_CTYPE=C tr -dc 'a-zA-Z0-9' | fold -w 64 | head -n 1) echo $SUBMARINER_PSK
-
Update the helm repository to pull the latest version of the Submariner charts
helm repo update
Each cluster that will be connected must have Submariner installed within it. You must repeat these steps for each cluster that you add.
-
Set your kubeconfig context to your desired installation cluster
kubectl config use-context <CLUSTER_CONTEXT>
-
Label your gateway nodes with the annotation
submariner.io/gateway=true
kubectl label node <DESIRED_NODE> "submariner.io/gateway=true"
-
Initialize Helm (if not yet done)
kubectl -n kube-system create serviceaccount tiller kubectl create clusterrolebinding tiller \ --clusterrole=cluster-admin \ --serviceaccount=kube-system:tiller helm init --service-account tiller
-
Wait for Tiller to initialize
kubectl -n kube-system rollout status deploy/tiller-deploy
-
Install submariner into this cluster. The values within the following command correspond to the table below.
helm install submariner-latest/submariner \ --name submariner \ --namespace submariner \ --set ipsec.psk="${SUBMARINER_PSK}" \ --set broker.server="${SUBMARINER_BROKER_URL}" \ --set broker.token="${SUBMARINER_BROKER_TOKEN}" \ --set broker.namespace="${SUBMARINER_BROKER_NS}" \ --set broker.ca="${SUBMARINER_BROKER_CA}" \ \ --set submariner.clusterId="<CLUSTER_ID>" \ --set submariner.clusterCidr="<CLUSTER_CIDR>" \ --set submariner.serviceCidr="<SERVICE_CIDR>" \ --set submariner.natEnabled="<NAT_ENABLED>"
Placeholder Description Default Example <CLUSTER_ID> Cluster ID (Must be RFC 1123 compliant) "" west-cluster <CLUSTER_CIDR> Cluster CIDR for Cluster "" 10.42.0.0/16
<SERVICE_CIDR> Service CIDR for Cluster "" 10.43.0.0/16
<NAT_ENABLED> If in a cloud provider that uses 1:1 NAT between instances (for example, AWS VPC), you should set this to true
so that Submariner is aware of the 1:1 NAT condition."false" false
Switch to the context of one of your clusters, i.e. kubectl config use-context west
Run an nginx container in this cluster, i.e. kubectl run -n default nginx --image=nginx
Retrieve the pod IP of the nginx container, looking under the "Pod IP" column for kubectl get pod -n default
Change contexts to your other workload cluster, i.e. kubectl config use-context east
Run a busybox pod and ping/curl the nginx pod:
kubectl run -i -t busybox --image=busybox --restart=Never
If you don't see a command prompt, try pressing enter.
/ # ping <NGINX_POD_IP>
/ # wget -O - <NGINX_POD_IP>
When running in AWS, it is necessary to disable source/dest checking of the instances that are gateway hosts to allow the instances to pass traffic for remote clusters.
To build submariner-engine
and submariner-route-agent
you can trigger make
, which will perform a Dapperized build of the components.
We welcome issues/PR's to Submariner, if you encounter issues that you'd like to fix while working on it or find new features that you'd like.
- Potentially spin out Charon into it's own pod to help decrease downtime
- Better clean up of IPtables rules when node loses leader election
- Central API server that is hosted by Rancher