调试 Kubernetes 中的 DNS 解析问题

lai*_*son 5 kubernetes ubuntu-18.04

我在 Ubuntu 18.04 上使用 Kubespray 构建了一个 Kubernetes 集群,并面临 DNS 问题,因此基本上容器无法通过其主机名进行通信。

正在发挥作用的事情:

  • 容器通过IP地址进行通信
  • 互联网正在容器中运行
  • 能够解决kubernetes.default

库伯内特大师:

root@k8s-1:~# cat /etc/resolv.conf | grep -v ^\\#
nameserver 127.0.0.53
search home
root@k8s-1:~# 
Run Code Online (Sandbox Code Playgroud)

荚:

root@k8s-1:~# kubectl exec dnsutils cat /etc/resolv.conf
nameserver 169.254.25.10
search default.svc.cluster.local svc.cluster.local cluster.local home
options ndots:5
root@k8s-1:~# 
Run Code Online (Sandbox Code Playgroud)

CoreDNS Pod 运行状况良好:

root@k8s-1:~# kubectl get pods --namespace=kube-system -l k8s-app=kube-dns        
NAME                       READY   STATUS    RESTARTS   AGE
coredns-58687784f9-8rmlw   1/1     Running   0          35m
coredns-58687784f9-hp8hp   1/1     Running   0          35m
root@k8s-1:~#
Run Code Online (Sandbox Code Playgroud)

CoreDNS Pod 的日志:

root@k8s-1:~# kubectl describe pods --namespace=kube-system -l k8s-app=kube-dns | tail -n 2
  Normal   Started           35m                 kubelet, k8s-2     Started container coredns
  Warning  DNSConfigForming  12s (x33 over 35m)  kubelet, k8s-2     Nameserver limits were exceeded, some nameservers have been omitted, the applied nameserver line is: 4.2.2.1 4.2.2.2 208.67.220.220

root@k8s-1:~# kubectl logs --namespace=kube-system coredns-58687784f9-8rmlw
.:53
2020-02-09T22:56:14.390Z [INFO] plugin/reload: Running configuration MD5 = b9d55fc86b311e1d1a0507440727efd2
2020-02-09T22:56:14.391Z [INFO] CoreDNS-1.6.0
2020-02-09T22:56:14.391Z [INFO] linux/amd64, go1.12.7, 0a218d3
CoreDNS-1.6.0
linux/amd64, go1.12.7, 0a218d3
root@k8s-1:~#

root@k8s-1:~# kubectl logs --namespace=kube-system coredns-58687784f9-hp8hp
.:53
2020-02-09T22:56:20.388Z [INFO] plugin/reload: Running configuration MD5 = b9d55fc86b311e1d1a0507440727efd2
2020-02-09T22:56:20.388Z [INFO] CoreDNS-1.6.0
2020-02-09T22:56:20.388Z [INFO] linux/amd64, go1.12.7, 0a218d3
CoreDNS-1.6.0
linux/amd64, go1.12.7, 0a218d3
root@k8s-1:~#
Run Code Online (Sandbox Code Playgroud)

CoreDNS 似乎暴露了:

root@k8s-1:~# kubectl get svc --namespace=kube-system | grep coredns
coredns                ClusterIP   10.233.0.3      <none>        53/UDP,53/TCP,9153/TCP   37m
root@k8s-1:~#

root@k8s-1:~# kubectl get ep coredns --namespace=kube-system
NAME      ENDPOINTS                                                  AGE
coredns   10.233.64.2:53,10.233.65.3:53,10.233.64.2:53 + 3 more...   37m
root@k8s-1:~#
Run Code Online (Sandbox Code Playgroud)

这些是我有问题的 pod - 所有集群都因此问题而受到影响:

root@k8s-1:~# kubectl get pods -o wide -n default
NAME                     READY   STATUS    RESTARTS   AGE   IP            NODE    NOMINATED NODE   READINESS GATES
busybox                  1/1     Running   0          17m   10.233.66.7   k8s-3   <none>           <none>
dnsutils                 1/1     Running   0          50m   10.233.66.5   k8s-3   <none>           <none>
nginx-86c57db685-p8zhc   1/1     Running   0          43m   10.233.64.3   k8s-1   <none>           <none>
nginx-86c57db685-st7rw   1/1     Running   0          47m   10.233.66.6   k8s-3   <none>           <none>
root@k8s-1:~# 
Run Code Online (Sandbox Code Playgroud)

能够使用 DNS 和容器通过 IP 地址访问互联网:

root@k8s-1:~# kubectl exec -it nginx-86c57db685-st7rw -- sh -c "ping 10.233.64.3"
PING 10.233.64.3 (10.233.64.3) 56(84) bytes of data.
64 bytes from 10.233.64.3: icmp_seq=1 ttl=62 time=0.481 ms
64 bytes from 10.233.64.3: icmp_seq=2 ttl=62 time=0.551 ms
...

root@k8s-1:~# kubectl exec -it nginx-86c57db685-st7rw -- sh -c "ping google.com"
PING google.com (172.217.21.174) 56(84) bytes of data.
64 bytes from fra07s64-in-f174.1e100.net (172.217.21.174): icmp_seq=1 ttl=61 time=77.9 ms
...

root@k8s-1:~# kubectl exec -it nginx-86c57db685-st7rw -- sh -c "ping kubernetes.default"
PING kubernetes.default.svc.cluster.local (10.233.0.1) 56(84) bytes of data.
64 bytes from kubernetes.default.svc.cluster.local (10.233.0.1): icmp_seq=1 ttl=64 time=0.030 ms
64 bytes from kubernetes.default.svc.cluster.local (10.233.0.1): icmp_seq=2 ttl=64 time=0.069 ms
...
Run Code Online (Sandbox Code Playgroud)

实际问题:

root@k8s-1:~# kubectl exec -it nginx-86c57db685-st7rw -- sh -c "ping nginx-86c57db685-p8zhc"
ping: nginx-86c57db685-p8zhc: Name or service not known
command terminated with exit code 2
root@k8s-1:~#

root@k8s-1:~# kubectl exec -it nginx-86c57db685-st7rw -- sh -c "ping dnsutils"
ping: dnsutils: Name or service not known
command terminated with exit code 2
root@k8s-1:~#

oot@k8s-1:~# kubectl exec -ti busybox -- nslookup nginx-86c57db685-p8zhc
Server:     169.254.25.10
Address:    169.254.25.10:53

** server can't find nginx-86c57db685-p8zhc.default.svc.cluster.local: NXDOMAIN

*** Can't find nginx-86c57db685-p8zhc.svc.cluster.local: No answer
*** Can't find nginx-86c57db685-p8zhc.cluster.local: No answer
*** Can't find nginx-86c57db685-p8zhc.home: No answer
*** Can't find nginx-86c57db685-p8zhc.default.svc.cluster.local: No answer
*** Can't find nginx-86c57db685-p8zhc.svc.cluster.local: No answer
*** Can't find nginx-86c57db685-p8zhc.cluster.local: No answer
*** Can't find nginx-86c57db685-p8zhc.home: No answer

command terminated with exit code 1
root@k8s-1:~#
Run Code Online (Sandbox Code Playgroud)

我是否遗漏了某些内容或如何使用主机名修复容器之间的通信?

非常感谢

更新

更多检查:

root@k8s-1:~# kubectl exec -ti dnsutils -- nslookup kubernetes.default
Server:     169.254.25.10
Address:    169.254.25.10#53

Name:   kubernetes.default.svc.cluster.local
Address: 10.233.0.1
Run Code Online (Sandbox Code Playgroud)

我创建了 StatefulSet:

kubectl apply -f https://raw.githubusercontent.com/kubernetes/website/master/content/en/examples/application/web/web.yaml
Run Code Online (Sandbox Code Playgroud)

我能够 ping 服务“nginx”:

root@k8s-1:~/kplay# k exec dnsutils -it nslookup nginx
Server:     169.254.25.10
Address:    169.254.25.10#53

Name:   nginx.default.svc.cluster.local
Address: 10.233.66.8
Name:   nginx.default.svc.cluster.local
Address: 10.233.64.3
Name:   nginx.default.svc.cluster.local
Address: 10.233.65.5
Name:   nginx.default.svc.cluster.local
Address: 10.233.66.6
Run Code Online (Sandbox Code Playgroud)

使用 FQDN 时还能够联系有状态集成员

root@k8s-1:~/kplay# k exec dnsutils -it nslookup web-0.nginx.default.svc.cluster.local
Server:     169.254.25.10
Address:    169.254.25.10#53

Name:   web-0.nginx.default.svc.cluster.local
Address: 10.233.65.5

root@k8s-1:~/kplay# k exec dnsutils -it nslookup web-1.nginx.default.svc.cluster.local
Server:     169.254.25.10
Address:    169.254.25.10#53

Name:   web-1.nginx.default.svc.cluster.local
Address: 10.233.66.8
Run Code Online (Sandbox Code Playgroud)

但不只使用主机名:

root@k8s-1:~/kplay# k exec dnsutils -it nslookup web-0
Server:     169.254.25.10
Address:    169.254.25.10#53

** server can't find web-0: NXDOMAIN

command terminated with exit code 1
root@k8s-1:~/kplay# k exec dnsutils -it nslookup web-1
Server:     169.254.25.10
Address:    169.254.25.10#53

** server can't find web-1: NXDOMAIN

command terminated with exit code 1
root@k8s-1:~/kplay#
Run Code Online (Sandbox Code Playgroud)

它们都生活在同一个命名空间中:

root@k8s-1:~/kplay# k get pods -n default
NAME                     READY   STATUS    RESTARTS   AGE
busybox                  1/1     Running   22         22h
dnsutils                 1/1     Running   22         22h
nginx-86c57db685-p8zhc   1/1     Running   0          22h
nginx-86c57db685-st7rw   1/1     Running   0          22h
web-0                    1/1     Running   0          11m
web-1                    1/1     Running   0          10m
Run Code Online (Sandbox Code Playgroud)

另一项测试确认我能够 ping 服务:

kubectl create deployment --image nginx some-nginx
kubectl scale deployment --replicas 2 some-nginx
kubectl expose deployment some-nginx --port=12345 --type=NodePort

root@k8s-1:~/kplay# k exec dnsutils -it nslookup some-nginx
Server:     169.254.25.10
Address:    169.254.25.10#53

Name:   some-nginx.default.svc.cluster.local
Address: 10.233.63.137
Run Code Online (Sandbox Code Playgroud)

最后的想法

有趣的是,但这也许就是 Kubernetes 应该如何工作的?如果我想单独访问某个 Pod,我可以访问服务主机名和有状态集成员。至少在我的 k8s 使用中(可能适合每个人),如果不是有状态集,则到达单个 pod 似乎不是很重要。

Mar*_*ney 2


我建议您遵循此操作,以便我们可以隔离您的 CoreDNS 中可能存在的问题,并且您可以看到它工作正常。

至少在我的 k8s 使用中(可能适合每个人),如果不是有状态集,则到达单个 pod 似乎不是很重要。

可以使用 DNS 记录访问 pod,但正如您所说,这对于常规 K8s 实现来说并不是很重要。

启用后,Pod 会被分配一条 DNS A 记录,格式为 pod-ip-address.my-namespace.pod.cluster.local.

例如, 1.2.3.4 命名空间中 具有 IP default 、DNS 名称为 的 podcluster.local 将具有一个条目: 1-2-3-4.default.pod.cluster.local来源

例子

$ kubectl get pods -o wide
NAME         READY   STATUS    RESTARTS   AGE     IP          NODE                                 NOMINATED NODE   READINESS GATES
dnsutils     1/1     Running   20         20h     10.28.2.3   gke-lab-default-pool-87c6b085-wcp8   <none>           <none>
sample-pod   1/1     Running   0          2m11s   10.28.2.4   gke-lab-default-pool-87c6b085-wcp8   <none>           <none>

$ kubectl exec -ti dnsutils -- nslookup 10-28-2-4.default.pod.cluster.local
Server:     10.31.240.10
Address:    10.31.240.10#53

Name:   10-28-2-4.default.pod.cluster.local
Address: 10.28.2.4
Run Code Online (Sandbox Code Playgroud)

有趣的是,但这也许就是 Kubernetes 应该如何工作的?

是的,您的 CoreDNS 正在按预期工作,并且您描述的一切都是预期的。