controller是什么
简单来说,controller就是监听kubernetes的配置/状态变化,自动对资源进行调节。比如kubernetes内置的controller有Deployment、StatefulSet、DaemonSet。
实现controller
这里我们实现一个简单的loadbalancer controller,使用kubernetes的client-go。
先创建一个新的Serivce,并且指定Type为LoadBalancer。
apiVersion: v1
kind: Service
metadata:
name: my-service
spec:
selector:
app.kubernetes.io/name: MyApp
ports:
- protocol: TCP
port: 80
targetPort: 80
clusterIP: 10.0.171.239
type: LoadBalancer
这时Service的状态还是pending,因为还没有controller为其分配IP地址。
创建新的client
config, err := rest.InClusterConfig()
if err != nil {
panic(err)
}
cli, err := kubernetes.NewForConfig(config)
if err != nil {
panic(err)
}
创建一个新的对Service状态变化的监听
wch, err := cli.CoreV1().Services(corev1.NamespaceAll).Watch(context.TODO(), metav1.ListOptions{Watch: true})
if err != nil {
panic(err)
}
defer wch.Stop()
如果是新建的Serivce,并且Type为LoadBalancer,则自动为其分配一个IP地址。
// 端口分配范围
var portRange = atomic.Uint32{}
func init() {
portRange.Store(30000)
}
for {
select {
case <-signChannel:
return
case obj, ok := <-wch.ResultChan():
if !ok {
return
}
svc, ok := obj.Object.(*corev1.Service)
if !ok {
continue
}
switch obj.Type {
case watch.Added:
// 判断是否是LoadBalancer
// 生成环境中还应结合Annotations进行判断,防止误更改错误的Service.
if svc.Spec.Type != corev1.ServiceTypeLoadBalancer {
continue
}
// 生成端口,并创建转发(只会打印,不会真正创建规则)
sport := portRange.Add(1)
forward(svc, sport)
// 为service分配IP地址
svc.Status.LoadBalancer.Ingress = []corev1.LoadBalancerIngress{
{
IP: svc.Spec.ClusterIP,
Ports: []corev1.PortStatus{
{
Port: int32(sport),
},
},
},
}
// 更新service
_, err := cli.
CoreV1().
Services(svc.Namespace).
UpdateStatus(context.TODO(), svc, metav1.UpdateOptions{})
if err != nil {
slog.Error(err.Error())
} else {
slog.Info("service added", svc.Name, svc.Namespace, svc.Spec.ClusterIP, svc.Spec.ClusterIPs)
}
case watch.Modified:
slog.Info("service modified", svc.Name, svc.Namespace, svc.Spec.ClusterIP, svc.Spec.ClusterIPs)
case watch.Deleted:
slog.Info("service deleted", svc.Name, svc.Namespace, svc.Spec.ClusterIP, svc.Spec.ClusterIPs)
}
}
}
// ...
// ...
// 模拟实现iptables自动创建转发
func forward(svc *corev1.Service, sport uint32) {
ip, err := netip.ParseAddr(svc.Spec.ClusterIP)
if err != nil {
panic(err)
}
for _, v := range svc.Spec.Ports {
var ipt string
var mask string
if ip.Is4() {
ipt = "iptables"
mask = "32"
} else {
ipt = "ip6tables"
mask = "128"
}
fmt.Println(
ipt, "-t", "filter", "-A", "FORWARD",
"-d", svc.Spec.ClusterIP+"/"+mask,
"-p", string(v.Protocol),
"--dport", strconv.Itoa(int(v.Port)),
"-j", "DROP",
)
fmt.Println(
ipt, "-t", "nat", "-I", "PRETROUTING", 0,
"-p", string(v.Protocol),
"--dport", strconv.Itoa(int(sport)),
"-j", "DNAT",
"--to", net.JoinHostPort(svc.Spec.ClusterIP, strconv.Itoa(int(v.Port))),
)
fmt.Println(
ipt, "-t", "nat", "-I", "POSTROUTING",
"-d", svc.Spec.ClusterIP+"/"+mask,
"-p", v.Protocol,
"-j", "MASQUERADE",
)
}
}
运行这个简单程序,再观察Serivce, 就会发现Serivice已经不再处于pending状态,并也为其分配了一个IP地址,虽然不能真正使用此地址进行访问,因为我们并没有创建真正的转发规则。
apiVersion: v1
kind: Service
metadata:
name: my-service
spec:
selector:
app.kubernetes.io/name: MyApp
ports:
- protocol: TCP
port: 80
targetPort: 80
clusterIP: 10.0.171.239
type: LoadBalancer
status:
loadBalancer:
ingress:
- ip: 10.0.171.239
k3s的klipper-lb是一个很简单loadbalancer实现,代码只是几行脚本。
start_proxy() {
for src_range in ${SRC_RANGES//,/ }; do
if echo ${src_range} | grep -Eq ":"; then
ip6tables -t filter -I FORWARD -s ${src_range} -p ${DEST_PROTO} --dport ${DEST_PORT} -j ACCEPT
else
iptables -t filter -I FORWARD -s ${src_range} -p ${DEST_PROTO} --dport ${DEST_PORT} -j ACCEPT
fi
done
for dest_ip in ${DEST_IPS//,/ }; do
if echo ${dest_ip} | grep -Eq ":"; then
[ $(cat /proc/sys/net/ipv6/conf/all/forwarding) == 1 ] || exit 1
ip6tables -t filter -A FORWARD -d ${dest_ip}/128 -p ${DEST_PROTO} --dport ${DEST_PORT} -j DROP
ip6tables -t nat -I PREROUTING -p ${DEST_PROTO} --dport ${SRC_PORT} -j DNAT --to [${dest_ip}]:${DEST_PORT}
ip6tables -t nat -I POSTROUTING -d ${dest_ip}/128 -p ${DEST_PROTO} -j MASQUERADE
else
[ $(cat /proc/sys/net/ipv4/ip_forward) == 1 ] || exit 1
iptables -t filter -A FORWARD -d ${dest_ip}/32 -p ${DEST_PROTO} --dport ${DEST_PORT} -j DROP
iptables -t nat -I PREROUTING -p ${DEST_PROTO} --dport ${SRC_PORT} -j DNAT --to ${dest_ip}:${DEST_PORT}
iptables -t nat -I POSTROUTING -d ${dest_ip}/32 -p ${DEST_PROTO} -j MASQUERADE
fi
done
}