adding support for HA clusters, using HAProxy (#317)

* adding support for HA clusters, using HAProxy

Signed-off-by: David Lyle <dklyle0@gmail.com>

* fixing typo

* fixing load balancer port default value

clr-k8s-examples/README.md

@@ -63,6 +63,28 @@ devicemapper storage. This should not be used for production.
 
 > NOTE: This step is done automatically if using vagrant.
 
+### For HA, setup the load balancer node
+
+Ideally, the load balancer node will be a separate node. However, one of the
+master nodes can also serve as the load balancer for the cluster. [HAProxy](https://www.haproxy.org/)
+is used in these instructions.
+
+```bash
+sudo swupd bundle-add haproxy
+sudo systemctl enable haproxy
+```
+
+Edit the master IP addresses and load balancer address and ports in [`haproxy.cfg.example`](haproxy.cfg.example)
+to match the IPs for the new cluster. If using a master node for the load balancer
+make sure that the `frontend bind` port is different than the Kubernetes API port, 6443.
+If using a separate machine for load balancing, the port can be 6443 if desired.
+
+```bash
+sudo mkdir -p /etc/haproxy
+sudo cp haproxy.cfg.example /etc/haproxy/haproxy.cfg
+sudo systemctl start haproxy
+```
+
 ## Bring up the master
 
 Run [`create_stack.sh`](create_stack.sh) on the master node. This sets up the
@@ -75,8 +97,9 @@ more information.
 
 > NOTE: Before running [`create_stack.sh`](create_stack.sh) script, make sure to export
 the necessary environment variables if needed to be changed. By default it will use
-`CLRK8S_CNI` to be canal, and `CLRK8S_RUNNER` to be crio. Cilium is tested only in the 
-Vagrant.
+`CLRK8S_CNI` to be canal, and `CLRK8S_RUNNER` to be crio. Cilium is tested only in the
+Vagrant. If creating an HA cluster, make sure to specify `LOAD_BALANCER_IP` and
+`LOAD_BALANCER_PORT`.
 
 ```bash
 # default shows help
@@ -86,6 +109,13 @@ Vagrant.
 In order to enable running greater than 110 pods per node, set the environment
 variable `HIGH_POD_COUNT` to any non-empty value.
 
+If creating an HA cluster, join the other master nodes to the cluster.
+
+```bash
+kubeadm join <load-balancer-ip>:<load-balancer-port> --token <token> --discovery-token-ca-cert-hash <hash> \
+    --control-plane --certificate-key <certificate-key> --cri-socket=/run/crio/crio.sock
+```
+
 ## Join Workers to the cluster
 
 ```bash
@@ -94,6 +124,9 @@ kubeadm join <master-ip>:<master-port> --token <token> --discovery-token-ca-cert
 
 Note: Remember to append `--cri-socket=/run/crio/crio.sock` to the join command generated by the master.
 
+If creating an HA cluster, join the other worker nodes to the cluster. The same way,
+but replacing the `<master-ip>:<master-port>` with `<load-balancer-ip>:<load-balancer-port>`.
+
 On workers just use the join command that the master spits out. There nothing
 else you need to run on the worker. All the other Kubernetes customizations are pushed
 in from master via the values setup in the `kubeadm.yaml` file.

clr-k8s-examples/create_stack.sh

@@ -14,6 +14,8 @@ SCRIPT_DIR="$(dirname "${BASH_SOURCE[0]}")"
 : ${MASTER_IP:=}
 : ${CERT_SANS:=}
 HIGH_POD_COUNT=${HIGH_POD_COUNT:-""}
+LOAD_BALANCER_IP=${LOAD_BALANCER_IP:-""}
+LOAD_BALANCER_PORT="${LOAD_BALANCER_PORT:-6444}"
 
 # versions
 CANAL_VER="${CLRK8S_CANAL_VER:-v3.10}"
@@ -90,7 +92,13 @@ function cluster_init() {
 		echo "/var/lib/etcd exists! skipping init."
 		return
 	fi
-	sudo -E kubeadm init --config=./kubeadm.yaml
+
+	if [[ -n "${LOAD_BALANCER_IP}" ]]; then
+		sed -i "s/ClusterConfiguration/ClusterConfiguration\ncontrolPlaneEndpoint: ${LOAD_BALANCER_IP}:${LOAD_BALANCER_PORT}/g" ./kubeadm.yaml
+	fi
+	# upload-certs will automatically upload certificates that should be shared
+	# across control-plane nodes in HA clusters. It is harmless in non-HA cases.
+	sudo -E kubeadm init --upload-certs --config=./kubeadm.yaml
 
 	rm -rf "${HOME}/.kube"
 	mkdir -p "${HOME}/.kube"

clr-k8s-examples/haproxy.cfg.example

@@ -0,0 +1,34 @@
+global
+    log /dev/log    local0
+    chroot /var/lib/haproxy
+    stats socket /run/haproxy-master.sock mode 660 level admin
+    stats timeout 30s
+    user haproxy
+    group haproxy
+    daemon
+    # Default SSL material locations
+    ca-base /etc/ssl/certs
+    ssl-default-bind-ciphers ECDH+AESGCM:DH+AESGCM:ECDH+AES256:DH+AES256:ECDH+AES128:DH+AES:RSA+AESGCM:RSA+AES:!aNULL:!MD5:!DSS
+    ssl-default-bind-options no-sslv3
+defaults
+    log     global
+    mode    http
+    option  httplog
+    option  dontlognull
+    timeout connect 5000
+    timeout client  50000
+    timeout server  50000
+    timeout tunnel  4h
+frontend kubernetes
+    bind 10.0.0.100:6444
+    option tcplog
+    mode tcp
+    default_backend kubernetes-master-nodes
+
+backend kubernetes-master-nodes
+    mode tcp
+    balance source
+    option tcp-check
+    server master-1 10.0.0.100:6443 check fall 3 rise 2
+    server master-2 10.0.0.101:6443 check fall 3 rise 2
+    server master-3 10.0.0.102:6443 check fall 3 rise 2

clr-k8s-examples/setup_system.sh

@@ -210,7 +210,7 @@ if [[ -n "${HIGH_POD_COUNT}" ]]; then
 fi
 echo "Reloading daemons..."
 daemon_reload
-echo "Enabling Kublet runner..."
+echo "Enabling kubelet runner..."
 enable_kubelet_runner
 echo "Ensuring system is ready..."
 ensure_system_ready