我在私有 GKE 对等连接和 Cloud NAT 方面做错了什么?

我在私有 GKE 对等连接和 Cloud NAT 方面做错了什么?

我正在尝试在 Google Cloud 上安装以下平台:

2 个私有(本机)GKE 集群,位于 2 个不同的 VPC 中,为了让它们可以访问互联网,每个 vpc 都配置了一个 Cloud Nat。

我需要的是 2 个 GKE 集群进行交互,但在对等 VPC 时,我只能获得 POD 之间的通信,而不是 POD -> 服务或 POD -> 内部负载均衡器之间的通信。

集群:

NAME     LOCATION      MASTER_VERSION  MASTER_IP        MACHINE_TYPE  NODE_VERSION   NUM_NODES  STATUS
Shrek01  asia-east1-a  1.16.8-gke.15   <none>           g1-small      1.16.8-gke.15  3          RUNNING
Shrek02  asia-east2-a  1.15.9-gke.24   <none>           g1-small      1.15.9-gke.24  3          RUNNING

vpcs:

NAME         SUBNET_MODE  BGP_ROUTING_MODE  IPV4_RANGE  GATEWAY_IPV4
Shrek01      CUSTOM       REGIONAL
Shrek02      CUSTOM       REGIONAL

子网:

NAME                REGION        NETWORK      RANGE
Shrek01             asia-east1    Shrek01      192.168.13.0/24
Shrek02             asia-east2    Shrek02      192.168.14.0/24

对等:

NAME                                     NETWORK      PEER_PROJECT                  PEER_NETWORK                            AUTO_CREATE_ROUTES  STATE   STATE_DETAILS
Shrek01-Shrek01-peering                  Shrek01      pocprod2-2019001              Shrek02                                 True                ACTIVE  [2020-05-16T14:29:57.864-07:00]: Connected.
Shrek02-Shrek01-peering                  Shrek02      pocprod2-2019001              Shrek01                                 True                ACTIVE  [2020-05-16T14:29:57.864-07:00]: Connected.

防火墙规则:

  • “Shrek01-peering-ingress”
{
  "allowed": [
    {
      "IPProtocol": "all"
    }
  ],
  "creationTimestamp": "2020-05-16T16:05:14.829-07:00",
  "description": "",
  "direction": "INGRESS",
  "disabled": false,
  "id": "6807007164648771397",
  "kind": "compute#firewall",
  "logConfig": {
    "enable": false
  },
  "name": "peering-ingress",
  "network": "https://www.googleapis.com/compute/v1/projects/pocprod2-2019001/global/networks/Shrek01",
  "priority": 1000,
  "selfLink": "https://www.googleapis.com/compute/v1/projects/pocprod2-2019001/global/firewalls/peering-ingress",
  "sourceRanges": [
    "192.168.14.0/24",
    "10.113.64.0/19",
    "10.213.64.0/19"
  ]
}
  • “Shrek02-peering-ingress”
{
  "allowed": [
    {
      "IPProtocol": "all"
    }
  ],
  "creationTimestamp": "2020-05-16T16:24:28.545-07:00",
  "description": "",
  "direction": "INGRESS",
  "disabled": false,
  "id": "7130188648920500419",
  "kind": "compute#firewall",
  "logConfig": {
    "enable": false
  },
  "name": "Shrek02-peering-ingress",
  "network": "https://www.googleapis.com/compute/v1/projects/pocprod2-2019001/global/networks/Shrek02",
  "priority": 1000,
  "selfLink": "https://www.googleapis.com/compute/v1/projects/pocprod2-2019001/global/firewalls/Shrek02-peering-ingress",
  "sourceRanges": [
    "192.168.13.0/24",
    "10.113.32.0/19",
    "10.213.32.0/19"
  ]
}

k8s Shrek01集群:

  • kubectl get svc -o wide
NAME         TYPE           CLUSTER-IP     EXTERNAL-IP    PORT(S)        AGE   SELECTOR
kubernetes   ClusterIP      10.213.32.1    <none>         443/TCP        85m   <none>
nginx        LoadBalancer   10.213.60.14   192.168.13.7   80:32612/TCP   92s   app=nginx
nginx-cip    ClusterIP      10.213.34.24   <none>         80/TCP         93s   app=nginx
nginx-np     NodePort       10.213.35.31   <none>         80:30444/TCP   92s   app=nginx
  • kubectl get pod -o wide
NAME                     READY   STATUS    RESTARTS   AGE     IP             NODE                                     NOMINATED NODE   READINESS GATES
nginx-64b4f9bb85-9sjcp   1/1     Running   0          3m34s   10.113.34.11   gke-Shrek01-default-pool-f9ecbfcc-dz9z   <none>           <none>
nginx-64b4f9bb85-l2bzd   1/1     Running   0          3m34s   10.113.32.5    gke-Shrek01-default-pool-f9ecbfcc-pdll   <none>           <none>
nginx-64b4f9bb85-xd7kw   1/1     Running   0          3m34s   10.113.33.9    gke-Shrek01-default-pool-f9ecbfcc-v67d   <none>           <none>
  • kubectl get nodes -o wide
NAME                                     STATUS   ROLES    AGE   VERSION          INTERNAL-IP    EXTERNAL-IP   OS-IMAGE                             KERNEL-VERSION   CONTAINER-RUNTIME
gke-Shrek01-default-pool-f9ecbfcc-dz9z   Ready    <none>   89m   v1.16.8-gke.15   192.168.13.4                 Container-Optimized OS from Google   4.19.109+        docker://19.3.1
gke-Shrek01-default-pool-f9ecbfcc-pdll   Ready    <none>   89m   v1.16.8-gke.15   192.168.13.2                 Container-Optimized OS from Google   4.19.109+        docker://19.3.1
gke-Shrek01-default-pool-f9ecbfcc-v67d   Ready    <none>   89m   v1.16.8-gke.15   192.168.13.3                 Container-Optimized OS from Google   4.19.109+        docker://19.3.1
  • 来自 Shrek02 吊舱的检查:
root@nginx-5c66c56f55-8jwv2:/# echo ${MY_POD_IP} 
10.113.66.9

# internal load balancer
root@nginx-5c66c56f55-8jwv2:/# nc -vz 192.168.13.7 80
192.168.13.7: inverse host lookup failed: Unknown host
(UNKNOWN) [192.168.13.7] 80 (?) : Connection timed out

# intarnal load balancer's Cluster IP
root@nginx-5c66c56f55-8jwv2:/# nc -vz 10.213.60.14 80
10.213.60.14: inverse host lookup failed: Unknown host
(UNKNOWN) [10.213.60.14] 80 (?) : Connection timed out

# ClusterIP
root@nginx-5c66c56f55-8jwv2:/# nc -vz 10.213.34.24 80
10.213.34.24: inverse host lookup failed: Unknown host
(UNKNOWN) [10.213.34.24] 80 (?) : Connection timed out

# NodePort 
root@nginx-5c66c56f55-8jwv2:/# nc -vz 10.213.35.31 80
10.213.35.31: inverse host lookup failed: Unknown host
(UNKNOWN) [10.213.35.31] 80 (?) : Connection timed out

# Pod IP
root@nginx-5c66c56f55-8jwv2:/# nc -vz 10.113.34.11 80
10.113.34.11: inverse host lookup failed: Unknown host
(UNKNOWN) [10.113.34.11] 80 (?) open

root@nginx-5c66c56f55-8jwv2:/# nc -vz 10.113.32.5 80
10.113.32.5: inverse host lookup failed: Unknown host
(UNKNOWN) [10.113.32.5] 80 (?) open

root@nginx-5c66c56f55-8jwv2:/# nc -vz 10.113.33.9 80
10.113.33.9: inverse host lookup failed: Unknown host
(UNKNOWN) [10.113.33.9] 80 (?) open

我是不是忘记了什么步骤?我没有发现错误。

答案1

  • 我能够使用类似的环境将 pod 连接到 pod、将 pod 连接到节点端口以及将端口连接到外部负载均衡器,但需要注意的是:

内部负载均衡器仅在以下情况下才允许在 VPC Peering 上使用:

  • 对等网络中的客户端虚拟机 (VM) 实例与内部负载均衡器位于同一区域
  • 您配置全球访问。配置全局访问后,对等 VPC 网络的任何区域的客户端 VM 实例都可以访问您的内部 TCP/UDP 负载均衡器。内部 HTTP(S) 负载均衡不支持全局访问。

为了在不同区域使用内部全球访问和 VPC 对等连接,您需要两个选择

$ gcloud compute forwarding-rules update <LB_NAME> \
   --region=<REGION> \
   --allow-global-access
  • 并使用以下命令进行检查:
gcloud compute forwarding-rules describe <LB_NAME> \
   --region=us-west1 \
   --format="get(name,region,allowGlobalAccess)"
  • 如果你所在地区有少量内部负载均衡,另一种简单的更改方法是通过 GCP 负载均衡页面单击此处访问
    • 在里面姓名列中,单击集群区域的内部 TCP/UDP 负载均衡器(单击后,您将看到子网名称,如下例所示):

在此处输入图片描述

  • 然后点击编辑
  • 点击前端配置
  • 点击铅笔编辑
  • 在下面全球访问, 选择 使能够
  • 点击完毕
  • 打开并点击更新更新规则。
    • 等到规则应用完成。
    • 改变之后,我就可以从《Shrek02》中瞄准《Shrek01》中的内部 LB。(我将在下面给出例子)。

笔记:
ClusterIP:在集群内部 IP 上公开服务。选择此值会使服务仅可从集群内部访问。因此您将不会被路由到外部访问。


再生产:

  • 我按照您的参数创建了两个 VPC:
$ gcloud container clusters list
NAME         LOCATION        MASTER_VERSION  MASTER_IP        MACHINE_TYPE   NODE_VERSION       NUM_NODES  STATUS
shrek01      europe-west1-b  1.16.8-gke.15   XX.XXX.XX.XXX    g1-small       1.16.8-gke.15      3          RUNNING
shrek02      europe-west2-b  1.15.9-gke.24   XXX.XXX.XX.XXX   g1-small       1.15.9-gke.24      3          RUNNING

$ gcloud compute networks subnets list
NAME     REGION                   NETWORK  RANGE
shrek01  europe-west1             shrek01  192.168.13.0/24
shrek02  europe-west2             shrek02  192.168.14.0/24

$ gcloud compute networks peerings list-routes sh1-sh2 --network=shrek01 --region europe-west1 --direction=INCOMING
DEST_RANGE       TYPE                  NEXT_HOP_REGION  PRIORITY  STATUS
192.168.14.0/24  SUBNET_PEERING_ROUTE  europe-west2     1000      accepted
10.229.0.0/20    SUBNET_PEERING_ROUTE  europe-west2     1000      accepted
10.36.0.0/14     SUBNET_PEERING_ROUTE  europe-west2     1000      accepted

$ gcloud compute networks peerings list-routes sh2-sh1 --network=shrek02 --region europe-west2 --direction=INCOMING
DEST_RANGE       TYPE                  NEXT_HOP_REGION  PRIORITY  STATUS
192.168.13.0/24  SUBNET_PEERING_ROUTE  europe-west1     1000      accepted
10.154.0.0/20    SUBNET_PEERING_ROUTE  europe-west1     1000      accepted
10.24.0.0/14     SUBNET_PEERING_ROUTE  europe-west1     1000      accepted

确保我的节点可以在 VPC 之间进行 Ping 后,我将使用此 yaml 测试入口和连接

  • hello-1.yaml
apiVersion: apps/v1
kind: Deployment
metadata:
 name: hello-1
spec:
 replicas: 3
 selector:
   matchLabels:
     app: hello-1
 template:
   metadata:
     labels:
       app: hello-1
   spec:
     containers:
     - name: hello-1
       image: gcr.io/google-samples/hello-app:1.0
       ports:
       - name: http
         containerPort: 8080
---
apiVersion: v1
kind: Service
metadata:
 name: hello-1-svc
spec:
 type: NodePort
 selector:
   app: hello-1
 ports:
   - protocol: TCP
     port: 80
     targetPort: 8080
  • hello-2.yaml
apiVersion: apps/v1
kind: Deployment
metadata:
 name: hello-2
spec:
 replicas: 3
 selector:
   matchLabels:
     app: hello-2
 template:
   metadata:
     labels:
       app: hello-2
   spec:
     containers:
     - name: hello-2
       image: gcr.io/google-samples/hello-app:2.0
       ports:
       - name: http
         containerPort: 8080
---
apiVersion: v1
kind: Service
metadata:
 name: hello-2-svc
spec:
 type: NodePort
 selector:
   app: hello-2
 ports:
   - protocol: TCP
     port: 80
     targetPort: 8080
  • hello-ingress.yaml
apiVersion: networking.k8s.io/v1beta1
kind: Ingress
metadata:
  name: hello-ingress
  annotations:
    kubernetes.io/ingress.class: nginx
spec:
  rules:
    - host:
      http:
        paths:
          - path: /
            backend:
              serviceName: hello-1-svc
              servicePort: 80
          - path: /v2
            backend:
              serviceName: hello-2-svc
              servicePort: 80
  • 查看 pod 名称和 pod 的 IP、以及节点 IP 和负载均衡器/nodeport 端口。
o$ kubectl get pods -o wide
NAME                       READY   STATUS    RESTARTS   AGE    IP           NODE                                     NOMINATED NODE   READINESS GATES
hello-1-84d5994678-dx8dv   1/1     Running   0          140m   10.24.0.9    gke-shrek01-default-pool-5ffc38d7-bz35   <none>           <none>
hello-1-84d5994678-t74mn   1/1     Running   0          14m    10.24.1.3    gke-shrek01-default-pool-5ffc38d7-70sk   <none>           <none>
hello-1-84d5994678-zq7t2   1/1     Running   0          14m    10.24.2.9    gke-shrek01-default-pool-5ffc38d7-zfj6   <none>           <none>
hello-2-5c4f554ccc-b8j6f   1/1     Running   0          140m   10.24.0.10   gke-shrek01-default-pool-5ffc38d7-bz35   <none>           <none>
hello-2-5c4f554ccc-km4ph   1/1     Running   0          13m    10.24.1.4    gke-shrek01-default-pool-5ffc38d7-70sk   <none>           <none>
hello-2-5c4f554ccc-z4f6n   1/1     Running   0          13m    10.24.2.10   gke-shrek01-default-pool-5ffc38d7-zfj6   <none>           <none>

$ ubectl get services
NAME          TYPE        CLUSTER-IP      EXTERNAL-IP   PORT(S)        AGE
hello-1-svc   NodePort    10.154.13.186   <none>        80:32186/TCP   140m
hello-2-svc   NodePort    10.154.4.214    <none>        80:32450/TCP   140m

$ kubectl get svc ingress-nginx-controller -n ingress-nginx
NAME                       TYPE           CLUSTER-IP      EXTERNAL-IP    PORT(S)                      AGE
ingress-nginx-controller   LoadBalancer   10.154.10.104   192.168.13.5   80:30112/TCP,443:32156/TCP   4h20m

$ kubectl get ingress
NAME            HOSTS   ADDRESS        PORTS   AGE
hello-ingress   *       192.168.13.5   80      98m

$ kubectl get nodes -o wide
NAME                                     STATUS   ROLES    AGE     VERSION          INTERNAL-IP    EXTERNAL-IP      OS-IMAGE                             KERNEL-VERSION   CONTAINER-RUNTIME
gke-shrek01-default-pool-5ffc38d7-70sk   Ready    <none>   2d19h   v1.16.8-gke.15   192.168.13.3   XX.XXX.XX.XXX    Container-Optimized OS from Google   4.19.109+        docker://19.3.1
gke-shrek01-default-pool-5ffc38d7-bz35   Ready    <none>   2d19h   v1.16.8-gke.15   192.168.13.2   XXX.XXX.XX.XXX   Container-Optimized OS from Google   4.19.109+        docker://19.3.1
gke-shrek01-default-pool-5ffc38d7-zfj6   Ready    <none>   2d19h   v1.16.8-gke.15   192.168.13.4   XX.XXX.X.XXX     Container-Optimized OS from Google   4.19.109+        docker://19.3.1

现在我将连接到shrek02集群,创建一个 pod 并安装curl

project@cloudshell:~$ kubectl run ubuntu --image=ubuntu -it -- /bin/bash
root@ubuntu:/# apt update
root@ubuntu:/# apt install curl
root@ubuntu:/# exit

project@cloudshell:~$ kubectl get pods -o wide
NAME     READY   STATUS    RESTARTS   AGE     IP          NODE                                     NOMINATED NODE   READINESS GATES
ubuntu   1/1     Running   1          2m51s   10.36.1.6   gke-shrek02-default-pool-a7a08ac8-0lrz   <none>           <none>
  • 您可以看到我们位于shrek02,现在让我们测试与资源的连接shrek01。请记住kube-dns仅在集群内部可用,因此我们将使用 IP 进行连接:
project@cloudshell:~$ kubectl exec -it ubuntu -- /bin/bash
###Hello-1 POD:
root@ubuntu:/# curl 10.24.0.9:8080
Hello, world!
Version: 1.0.0
Hostname: hello-1-84d5994678-dx8dv
###Hello-2 POD:
root@ubuntu:/# curl 10.24.1.4:8080
Hello, world!
Version: 2.0.0
Hostname: hello-2-5c4f554ccc-km4ph

### HELLO-1-SVC USING NODE IP + NODEPORT:
root@ubuntu:/# curl 192.168.13.3:32186
Hello, world!
Version: 1.0.0
Hostname: hello-1-84d5994678-t74mn
### HELLO-2-SVC USING ANOTHER NODE IP + NODEPORT:
root@ubuntu:/# curl 192.168.13.2:32450
Hello, world!
Version: 2.0.0
Hostname: hello-2-5c4f554ccc-km4ph

### NOW LET'S TEST OUR INGRESS which routes "/" to hello-1 and "/v2" to hello-2:
root@ubuntu:/# curl 192.168.13.5/
Hello, world!
Version: 1.0.0
Hostname: hello-1-84d5994678-dx8dv

root@ubuntu:/# curl 192.168.13.5/v2 
Hello, world!
Version: 2.0.0
Hostname: hello-2-5c4f554ccc-b8j6f

我希望它能帮助您解决您的环境问题,如果您有任何问题,请在评论中告诉我。

相关内容