我正在运行版本为1.12的香草EKS Kubernetes。
我使用CNI Genie允许自定义选择pod在启动时使用的CNI,并且我已经安装了标准Calico CNI设置。
使用CNI Genie,我将默认CNI配置为AWS CNI(aws节点),并且所有Pod都照常启动,并从我的VPC子网中分配了IP。
然后,我选择性地将印花布用作我要测试的一些基本吊舱的CNI。我正在使用默认的calico 192.168.0.0/16 CIDR范围。如果Pod位于相同的EKS工作程序节点上,则一切正常。
核心DNS的运行也很好(只要我让coredns容器在aws CNI上运行即可)。
但是,如果吊舱移动到另一个工作程序节点,则它们之间的联网将无法在集群内部工作。
我已经检查了calico自动配置的工作节点上的路由表,这对我来说似乎很合理。
这是我在所有命名空间中列出的广泛的pod:
NAMESPACE NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE
default hello-node1-865588ccd7-64p5x 1/1 Running 0 31m 192.168.106.129 ip-10-0-2-31.eu-west-2.compute.internal <none>
default hello-node2-dc7bbcb74-gqpwq 1/1 Running 0 17m 192.168.25.193 ip-10-0-3-222.eu-west-2.compute.internal <none>
kube-system aws-node-cm2dp 1/1 Running 0 26m 10.0.3.222 ip-10-0-3-222.eu-west-2.compute.internal <none>
kube-system aws-node-vvvww 1/1 Running 0 31m 10.0.2.31 ip-10-0-2-31.eu-west-2.compute.internal <none>
kube-system calico-kube-controllers-56bfccb786-fc2j4 1/1 Running 0 30m 10.0.2.41 ip-10-0-2-31.eu-west-2.compute.internal <none>
kube-system calico-node-flmnl 1/1 Running 0 31m 10.0.2.31 ip-10-0-2-31.eu-west-2.compute.internal <none>
kube-system calico-node-hcmqd 1/1 Running 0 26m 10.0.3.222 ip-10-0-3-222.eu-west-2.compute.internal <none>
kube-system coredns-6c64c9f456-g2h9k 1/1 Running 0 30m 10.0.2.204 ip-10-0-2-31.eu-west-2.compute.internal <none>
kube-system coredns-6c64c9f456-g5lhl 1/1 Running 0 30m 10.0.2.200 ip-10-0-2-31.eu-west-2.compute.internal <none>
kube-system genie-plugin-hspts 1/1 Running 0 26m 10.0.3.222 ip-10-0-3-222.eu-west-2.compute.internal <none>
kube-system genie-plugin-vqd2d 1/1 Running 0 31m 10.0.2.31 ip-10-0-2-31.eu-west-2.compute.internal <none>
kube-system kube-proxy-jm7f7 1/1 Running 0 26m 10.0.3.222 ip-10-0-3-222.eu-west-2.compute.internal <none>
kube-system kube-proxy-nnp76 1/1 Running 0 31m 10.0.2.31 ip-10-0-2-31.eu-west-2.compute.internal <none>
如您所见,两个 hello-node 窗格使用的是 Calico CNI。
我为hello-node pod提供了两项服务:
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
hello-node1 ClusterIP 172.20.90.83 <none> 8081/TCP 43m
hello-node2 ClusterIP 172.20.242.22 <none> 8082/TCP 43m
我已经确认是否使用aws CNI启动hello节点pod,当它们使用群集服务名称在单独的主机上运行时,我可以在它们之间ping / curl。
当我如上所述使用Calico CNI时,一切停止工作。
在此测试群集中,我只有两个EKS worker主机。这是每个的路由:
K8s Worker 1路线
[ec2-user@ip-10-0-3-222 ~]$ ip route
default via 10.0.3.1 dev eth0
10.0.3.0/24 dev eth0 proto kernel scope link src 10.0.3.222
169.254.169.254 dev eth0
blackhole 192.168.25.192/26 proto bird
192.168.25.193 dev calia0da7d91dc2 scope link
192.168.106.128/26 via 10.0.2.31 dev tunl0 proto bird onlink
K8s Worker 2路线
[ec2-user@ip-10-0-2-31 ~]$ ip route
default via 10.0.2.1 dev eth0
10.0.2.0/24 dev eth0 proto kernel scope link src 10.0.2.31
10.0.2.41 dev enif4cf9019f11 scope link
10.0.2.200 dev eni412af1a0e55 scope link
10.0.2.204 dev eni04260ebbbe1 scope link
169.254.169.254 dev eth0
192.168.25.192/26 via 10.0.3.222 dev tunl0 proto bird onlink
blackhole 192.168.106.128/26 proto bird
192.168.106.129 dev cali19da7817849 scope link
对我来说,路线:
192.168.25.192/26 via 10.0.3.222 dev tunl0 proto bird onlink
告诉我,从该工作服务器(及其容器/吊舱)发往192.168.25.192/16子网的流量应在tunl0接口上发送至10.0.3.222(用于EC2主机的AWS VPC ENI)。
此路由位于EC2主机10.0.2.31
上。因此,换句话说,当从该主机的容器与calico子网192.168.25.192/16上的容器进行通信时,网络流量应路由至10.0.3.222(我的其他EKS工作节点的ENI IP,在该节点上,使用Calico的容器在该子网上运行)。
要阐明我的测试过程,请执行以下操作:
curl http://hello-node2:8082
(或ping hello-node2 窗格中的calico分配的IP地址。编辑
为了进一步测试,我在运行 hello-node2 容器的主机上运行tcpdump,捕获了端口8080(容器在此端口上侦听)。
我确实在要卷曲的测试容器正在运行的目标主机上获得活动,但是它似乎并不表示流量下降。
[ec2-user@ip-10-0-3-222 ~]$ sudo tcpdump -vv -x -X -i tunl0 'port 8080'
tcpdump: listening on tunl0, link-type RAW (Raw IP), capture size 262144 bytes
14:32:42.859238 IP (tos 0x0, ttl 254, id 63813, offset 0, flags [DF], proto TCP (6), length 60)
10.0.2.31.29192 > 192.168.25.193.webcache: Flags [S], cksum 0xf932 (correct), seq 3206263598, win 28000, options [mss 1400,sackOK,TS val 2836614698 ecr 0,nop,wscale 7], length 0
0x0000: 4500 003c f945 4000 fe06 9ced 0a00 021f E..<.E@.........
0x0010: c0a8 19c1 7208 1f90 bf1b b32e 0000 0000 ....r...........
0x0020: a002 6d60 f932 0000 0204 0578 0402 080a ..m`.2.....x....
0x0030: a913 4e2a 0000 0000 0103 0307 ..N*........
14:32:43.870168 IP (tos 0x0, ttl 254, id 63814, offset 0, flags [DF], proto TCP (6), length 60)
10.0.2.31.29192 > 192.168.25.193.webcache: Flags [S], cksum 0xf53f (correct), seq 3206263598, win 28000, options [mss 1400,sackOK,TS val 2836615709 ecr 0,nop,wscale 7], length 0
0x0000: 4500 003c f946 4000 fe06 9cec 0a00 021f E..<.F@.........
0x0010: c0a8 19c1 7208 1f90 bf1b b32e 0000 0000 ....r...........
0x0020: a002 6d60 f53f 0000 0204 0578 0402 080a ..m`.?.....x....
0x0030: a913 521d 0000 0000 0103 0307 ..R.........
^C
2 packets captured
2 packets received by filter
0 packets dropped by kernel
运行我的目标/测试容器的主机上的 calia0da7d91dc2 界面即使在我从其他主机上的另一个容器运行curl时,也会显示增加的RX数据包和字节数。流量肯定在穿越。
[ec2-user@ip-10-0-3-222 ~]$ ifconfig
calia0da7d91dc2: flags=4163<UP,BROADCAST,RUNNING,MULTICAST> mtu 1440
inet6 fe80::ecee:eeff:feee:eeee prefixlen 64 scopeid 0x20<link>
ether ee:ee:ee:ee:ee:ee txqueuelen 0 (Ethernet)
RX packets 84 bytes 5088 (4.9 KiB)
RX errors 0 dropped 0 overruns 0 frame 0
TX packets 0 bytes 0 (0.0 B)
TX errors 0 dropped 0 overruns 0 carrier 0 collisions 0
是什么原因导致此处的主机之间无法联网?我缺少明显的东西吗?
编辑2 -有关Arjun Pandey- parjun8840的信息
以下是有关Calico配置的更多信息:
calicoctl配置注意:似乎工作负载端点不存在...
me@mine ~ aws-vault exec my-vault-entry -- kubectl get IPPool --all-namespaces
NAME AGE
default-ipv4-ippool 1d
me@mine ~ aws-vault exec my-vault-entry -- kubectl get IPPool default-ipv4-ippool -o yaml
apiVersion: crd.projectcalico.org/v1
kind: IPPool
metadata:
annotations:
projectcalico.org/metadata: '{"uid":"41bd2c82-d576-11e9-b1ef-121f3d7b4d4e","creationTimestamp":"2019-09-12T15:59:09Z"}'
creationTimestamp: "2019-09-12T15:59:09Z"
generation: 1
name: default-ipv4-ippool
resourceVersion: "500448"
selfLink: /apis/crd.projectcalico.org/v1/ippools/default-ipv4-ippool
uid: 41bd2c82-d576-11e9-b1ef-121f3d7b4d4e
spec:
blockSize: 26
cidr: 192.168.0.0/16
ipipMode: CrossSubnet
natOutgoing: true
nodeSelector: all()
vxlanMode: Never
me@mine ~ aws-vault exec my-vault-entry -- calicoctl get nodes
NAME
ip-10-254-109-184.ec2.internal
ip-10-254-109-237.ec2.internal
ip-10-254-111-147.ec2.internal
me@mine ~ aws-vault exec my-vault-entry -- calicoctl get workloadendpoints
WORKLOAD NODE NETWORKS INTERFACE
me@mine ~
以下是有关群集中的示例主机和测试容器的容器网络之一的一些网络信息:
主机ip a
[ec2-user@ip-10-254-109-184 ~]$ ip a
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN group default qlen 1000
link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
inet 127.0.0.1/8 scope host lo
valid_lft forever preferred_lft forever
inet6 ::1/128 scope host
valid_lft forever preferred_lft forever
2: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 9001 qdisc mq state UP group default qlen 1000
link/ether 02:1b:79:d1:c5:bc brd ff:ff:ff:ff:ff:ff
inet 10.254.109.184/26 brd 10.254.109.191 scope global dynamic eth0
valid_lft 2881sec preferred_lft 2881sec
inet6 fe80::1b:79ff:fed1:c5bc/64 scope link
valid_lft forever preferred_lft forever
3: eni808caba7453@if4: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 9001 qdisc noqueue state UP group default
link/ether c2:be:80:d4:6a:f3 brd ff:ff:ff:ff:ff:ff link-netnsid 0
inet6 fe80::c0be:80ff:fed4:6af3/64 scope link
valid_lft forever preferred_lft forever
5: tunl0@NONE: <NOARP,UP,LOWER_UP> mtu 1440 qdisc noqueue state UNKNOWN group default qlen 1000
link/ipip 0.0.0.0 brd 0.0.0.0
inet 192.168.29.128/32 brd 192.168.29.128 scope global tunl0
valid_lft forever preferred_lft forever
6: eth1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 9001 qdisc mq state UP group default qlen 1000
link/ether 02:12:58:bb:c6:1a brd ff:ff:ff:ff:ff:ff
inet 10.254.109.137/26 brd 10.254.109.191 scope global eth1
valid_lft forever preferred_lft forever
inet6 fe80::12:58ff:febb:c61a/64 scope link
valid_lft forever preferred_lft forever
7: enia6f1918d9e2@if4: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 9001 qdisc noqueue state UP group default
link/ether 96:f5:36:53:e9:55 brd ff:ff:ff:ff:ff:ff link-netnsid 1
inet6 fe80::94f5:36ff:fe53:e955/64 scope link
valid_lft forever preferred_lft forever
8: enia32d23ac2d1@if4: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 9001 qdisc noqueue state UP group default
link/ether 36:5e:34:a7:82:30 brd ff:ff:ff:ff:ff:ff link-netnsid 2
inet6 fe80::345e:34ff:fea7:8230/64 scope link
valid_lft forever preferred_lft forever
9: cali5e7dde1e39e@if4: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1440 qdisc noqueue state UP group default
link/ether ee:ee:ee:ee:ee:ee brd ff:ff:ff:ff:ff:ff link-netnsid 3
inet6 fe80::ecee:eeff:feee:eeee/64 scope link
valid_lft forever preferred_lft forever
[ec2-user@ip-10-254-109-184 ~]$
在测试容器pid上输入以获取ip a
信息:
[ec2-user@ip-10-254-109-184 ~]$ sudo nsenter -t 15715 -n ip a
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN group default qlen 1000
link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
inet 127.0.0.1/8 scope host lo
valid_lft forever preferred_lft forever
2: tunl0@NONE: <NOARP> mtu 1480 qdisc noop state DOWN group default qlen 1000
link/ipip 0.0.0.0 brd 0.0.0.0
4: eth0@if9: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1440 qdisc noqueue state UP group default
link/ether 9a:6d:db:06:74:cb brd ff:ff:ff:ff:ff:ff link-netnsid 0
inet 192.168.29.129/32 scope global eth0
valid_lft forever preferred_lft forever
答案 0 :(得分:0)
我现在不确定确切的解决方案(我尚未在AWS上测试calico,通常我在AWS和物理集群calico上使用amazon-vpc-cni-k8s),但是以下是我们可以快速完成的工作调查。
Calico AWS要求-https://docs.projectcalico.org/v2.3/reference/public-cloud/aws
kubectl get IPPool --all-namespaces
NAME AGE
default-ipv4-ippool 15d
kubectl get IPPool default-ipv4-ippool -o yaml
~ calicoctl get nodes
NAME
node1
node2
node3
node4
~ calicoctl get workloadendpoints
NODE ORCHESTRATOR WORKLOAD NAME
node2 k8s default.myapp-569c54f85-xtktk eth0
node1 k8s kube-system.calico-kube-controllers-5cbcccc885-b9x8s eth0
node1 k8s kube-system.coredns-fb8b8dcde-2zpw8 eth0
node1 k8s kube-system.coredns-fb8b8dcfg-hc6zv eth0
此外,如果我们可以获得容器网络的详细信息: nsenter -t pid -n ip a
对于主机也是如此: ip a