2021-12-28 ecd059bd9f6423145d7848b3fdcb7164 99+ 1 小时 9.7 k0次访问

关于 kubernetes中网络CNI方面的一些笔记

认定一件事，即使拿十分力气都无法完成，也要拿出十二分力气去努力——烽火戏诸侯《剑来

写在前面

学习k8s遇到整理笔记
博文内容主要涉及
K8s网络理论体系简述
K8s中Calico的实现方案
K8s中Calico网络方案容器跨主机通信过程Demo
K8s中网络策略方式：egress和ingress的Demo
ipBlock,namespaceSelector,podSelector的网络策略规则Demo

认定一件事，即使拿十分力气都无法完成，也要拿出十二分力气去努力——烽火戏诸侯《剑来》

Kubernetes网络

K8s中Calico的实现方案

Calico的实现方案

Calico的核心组件包括: Felix, etcd, BGP Client (BIRD)、 BGPRoute Reflector

Felix,即Calico代理, “跑”在Kubernetes的Node节点上,主要负责配置路由及ACL等信息来确保Endpoint的连通状态。

etcd,分布式键值存储,主要负责网络元数据一致性,确保Calico网络状态的准确性,可以与Kubernetes共用。
BGP Client (BIRD),主要负责把Felix写入Kernel的路由信息分发到当前Calico网络,确保workload间通信的有效性。

BGP Route Reflector,大规模部署时使用,摒弃所有节点互联的Mesh模式,通过一个或者多个BGP Route Reflector来完成集中式路由分发。

将整个互联网的可扩展IP网络原则压缩到数据中心级别, Calico在每一个计算节点利用Linux Kernel实现了一个高效的vRouter来负责数据转发,而每个vRouter通过BGP协议把在其上运行的容器的路由信息向整个Calico网络内传播,小规模部署可以直接互联,大规模下可通过指定的BGP Route Reflector来完成。这样保证最终所有的容器间的数据流量都是通过IP包的方式完成互联的。

基于三层实现通信，在二层上没有任何加密包装，因此只能在私有的可靠网络上使用。

流量隔离基于iptables实现，并且从etcd中获取需要生成的隔离规则，因此会有一些性能上的隐患。

每个主机上都部署了Calico-Node作为虚拟路由器,并且可以通过Calico将宿主机组织成任意的拓扑集群。当集群中的容器需要与外界通信时,就可以通过BGP协议将网关物理路由器加入到集群中,使外界可以直接访问容器IP,而不需要做任何NAT之类的复杂操作。

K8s整体流程图

跨主机Docker网络通信

常见的跨主机通信方案主要有以下几种:

形式	描述
Host模式	容器直接使用宿主机的网络,这样天生就可以支持跨主机通信。这种方式虽然可以解决跨主机通信问题,但应用场景很有限,容易出现端口冲突,也无法做到隔离网络环境,一个容器崩溃很可能引起整个宿主机的崩溃。
端口绑定	通过绑定容器端口到宿主机端口,跨主机通信时使用“主机IP+端口的方式访问容器中的服务。显然,这种方式仅能支持网络栈的4层及以上的应用,·并且容器与宿主机紧耦合,很难灵活地处理问题,可扩展性不佳。
定义容器网络	使用`Open vSwitch`或`Flannel`等第三方`SDN`工具,为容器构建可以跨主机通信的网络环境。这类方案一般要求各个主机上的Dockero网桥的`cidr`不同,以避免出现IP冲突的问题,限制容器在宿主机上可获取的IP范围。并且在容器需要对集群外提供服务时,需要比较复杂的配置,对部署实施人员的网络技能要求比较高。

容器网络发展到现在,形成了两大阵营:

Docker的CNM;
Google, Coreos，Kuberenetes主导的CNI

CNM和CNI是网络规范或者网络体系,并不是网络实现因此并不关心容器网络的实现方式( Flannel或者Calico等), CNM和CNI关心的只是网络管理。

网络类型	描述
CNM (Container Network Model)	CNM的优势在于原生,容器网络和Docker容器,生命周期结合紧密;缺点是被Docker “绑架”。支持CNM网络规范的容器网络实现包括:Docker Swarm overlay, Macvlan & IP networkdrivers, Calico, Contiv, Weave等。
CNI ( Container Network Interface)	CNI的优势是兼容其他容器技术(如rkt)及上层编排系统(Kubernetes&Mesos),而且社区活跃势头迅猛;缺点是非Docker原生。支持CNI网络规范的容器网络实现包括: Kubernetes、 Weave,Macvlan, Calico, Flannel, Contiv.Mesos CNI等。

但从网络实现角度,又可分为:

网络实现角度描述

隧道方案 隧道方案在laas层的网络中应用也比较多,它的主要缺点是随着节点规模的增长复杂度会提升,而且出了网络问题后跟踪起来比较麻烦,大规模集群情况下这是需要考虑的一个问题

路由方案 一般是基于3层或者2层实现网络隔离和跨主机容器互通的,出了问题也很容易排查。
Calico :基于BGP协议的路由方案,支持很细致的ACL控制,对混合云亲和度比较高。
Macvlan:从逻辑和Kernel层来看,是隔离性和性能最优的方案。基于二层隔离,所以需要一层路由器支持,大多数云服务商不支持,所以混合云上比较难以实现。

网络实现角度	描述
隧道方案	隧道方案在laas层的网络中应用也比较多,它的主要缺点是随着节点规模的增长复杂度会提升,而且出了网络问题后跟踪起来比较麻烦,大规模集群情况下这是需要考虑的一个问题
路由方案	一般是`基于3层或者2层实现网络隔离和跨主机容器互通的`,出了问题也很容易排查。 `Calico` :基于BGP协议的路由方案,支持很细致的ACL控制,对混合云亲和度比较高。 `Macvlan`:从逻辑和Kernel层来看,是隔离性和性能最优的方案。基于二层隔离,所以需要一层路由器支持,大多数云服务商不支持,所以混合云上比较难以实现。

calico通信过程

Calico把每个操作系统的协议栈当作一个路由器,认为所有的容器是连在这个路由器上的网络终端,在路由器之间运行标准的路由协议-BGP,然后让它们自己去学习这个网络拓扑该如何转发。

Calico方案其实是一个纯三层的方案,也就是说让 **每台机器的协议栈的三层去确保两个容器、跨主机容器之间的三层连通性**。其网络模型如图所示。

网络模型

对于控制平面,其每个Calico节点上会运行两个主要的程序

程序	描述
一个是`Felix`	它会监听`etcd`,并从`etcd`获取事件,如该节点新增容器或者增加IP地址等。当在这个节点上创建出一个容器,并将其`网卡、IP, MAC`都设置好后,`Felix在内核的路由表`里面写一条数据,注明这个`IP应该配置到这张网卡`。
一个`标准的路由程序`	,它会从`内核`里面获取哪一些`IP的路由`发生了变化,然后通过`标准BGP的路由协议扩散`到整个其他宿主机上,通知外界这个IP在这里。

由于Calico是一种纯三层(网络层)的实现,因此可以避免与二层方案相关的数据包封装的操作,·中间没有任何的NAT,没有任何的Overlay,所以它的转发效率可能是所有方案中最高的。因为它的包直接走原生TCP/IP的协议栈,它的隔离也因为这个栈而变得好做。因为TCP/IP的协议栈提供了一整套的防火墙规则,所以它可以通过iptables的规则达到比较复杂的隔离逻辑。

Calico实现方案

拓扑模式

环境准备

这里我们通过calico来进行跨主机容器网络通信过程演示,
ansible网络测试

┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$ansible etcd -m ping
192.168.26.101 | SUCCESS => {
    "ansible_facts": {
        "discovered_interpreter_python": "/usr/bin/python"
    },
    "changed": false,
    "ping": "pong"
}
192.168.26.102 | SUCCESS => {
    "ansible_facts": {
        "discovered_interpreter_python": "/usr/bin/python"
    },
    "changed": false,
    "ping": "pong"
}
192.168.26.100 | SUCCESS => {
    "ansible_facts": {
        "discovered_interpreter_python": "/usr/bin/python"
    },
    "changed": false,
    "ping": "pong"
}
┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$

etcd集群测试，这里我们已经搭建好一个etcd集群，etcdctl member list查看集群列表

┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$ansible etcd -m shell -a "etcdctl member list"
192.168.26.102 | CHANGED | rc=0 >>
6f2038a018db1103, started, etcd-100, http://192.168.26.100:2380, http://192.168.26.100:2379,http://localhost:2379
bd330576bb637f25, started, etcd-101, http://192.168.26.101:2380, http://192.168.26.101:2379,http://localhost:2379
fbd8a96cbf1c004d, started, etcd-102, http://192.168.26.102:2380, http://192.168.26.100:2379,http://localhost:2379
192.168.26.101 | CHANGED | rc=0 >>
6f2038a018db1103, started, etcd-100, http://192.168.26.100:2380, http://192.168.26.100:2379,http://localhost:2379
bd330576bb637f25, started, etcd-101, http://192.168.26.101:2380, http://192.168.26.101:2379,http://localhost:2379
fbd8a96cbf1c004d, started, etcd-102, http://192.168.26.102:2380, http://192.168.26.100:2379,http://localhost:2379
192.168.26.100 | CHANGED | rc=0 >>
6f2038a018db1103, started, etcd-100, http://192.168.26.100:2380, http://192.168.26.100:2379,http://localhost:2379
bd330576bb637f25, started, etcd-101, http://192.168.26.101:2380, http://192.168.26.101:2379,http://localhost:2379
fbd8a96cbf1c004d, started, etcd-102, http://192.168.26.102:2380, http://192.168.26.100:2379,http://localhost:2379
┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$

docker安装启动,修改数据存储位置

┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$ansible etcd -m shell -a "yum -y install docker-ce"
┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$ansible etcd -m shell -a "systemctl enable docker --now"
┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$ansible etcd -m shell -a "systemctl status docker"
192.168.26.100 | CHANGED | rc=0 >>
● docker.service - Docker Application Container Engine
   Loaded: loaded (/usr/lib/systemd/system/docker.service; enabled; vendor preset: disabled)
   Active: active (running) since Sat 2022-01-01 20:27:17 CST; 10min ago
     Docs: https://docs.docker.com
     ...

修改docker启动参数:数据存储位置--cluster-store=

┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$ansible etcd -m shell -a "cat /usr/lib/systemd/system/docker.service | grep containerd.sock"
192.168.26.100 | CHANGED | rc=0 >>
ExecStart=/usr/bin/dockerd -H fd:// --containerd=/run/containerd/containerd.sock
192.168.26.102 | CHANGED | rc=0 >>
ExecStart=/usr/bin/dockerd -H fd:// --containerd=/run/containerd/containerd.sock
192.168.26.101 | CHANGED | rc=0 >>
ExecStart=/usr/bin/dockerd -H fd:// --containerd=/run/containerd/containerd.sock

这里我们直接使用SED来修改

┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$ansible 192.168.26.100 -m shell -a "sed -i 's#containerd\.sock#containerd.sock  --cluster-store=etcd
://192.168.26.100:2379#' /usr/lib/systemd/system/docker.service "
192.168.26.100 | CHANGED | rc=0 >>

┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$ansible 192.168.26.101 -m shell -a "sed -i 's#containerd\.sock#containerd.sock  --cluster-store=etcd://192.168.26.101:2379#' /usr/lib/systemd/system/docker.service "
192.168.26.101 | CHANGED | rc=0 >>

┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$ansible 192.168.26.102 -m shell -a "sed -i 's#containerd\.sock#containerd.sock  --cluster-store=etcd
://192.168.26.102:2379#' /usr/lib/systemd/system/docker.service "
192.168.26.102 | CHANGED | rc=0 >>

┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$

刷新Service文件，重启docker

┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$ansible etcd -m shell -a "systemctl daemon-reload; systemctl restart docker"
192.168.26.100 | CHANGED | rc=0 >>

192.168.26.102 | CHANGED | rc=0 >>

192.168.26.101 | CHANGED | rc=0 >>

┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$ansible etcd -m shell -a "systemctl status docker"

然后我们需要创建calico配置文件,这里我们通过ansilbe 的方式
使用file模块新建文件夹mkdir /etc/calico

1 2	┌──[root@vms81.liruilongs.github.io]-[~/ansible] └─$ansible etcd -m file -a "path=/etc/calico/ state=directory force=yes"

使用template模块创建配置文件

新建模板，这里使用到j2模板，魔法变量

┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$cat calicoctl.j2
apiVersion: v1
kind: calicoApiConfig
metadata:
spec:
  datastoreType: "etcdv2"
  etcdEndpoints: "http://{{inventory_hostname}}:2379"
┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$

calico集群创建配置文件

1 2	┌──[root@vms81.liruilongs.github.io]-[~/ansible] └─$ansible etcd -m template -a "src=calicoctl.j2 dest=/etc/calico/calicoctl.cfg force=yes"

核对创建的配置文件

┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$ansible etcd -m shell -a "cat /etc/calico/calicoctl.cfg"
192.168.26.100 | CHANGED | rc=0 >>
apiVersion: v1
kind: calicoApiConfig
metadata:
spec:
  datastoreType: "etcdv2"
  etcdEndpoints: "http://192.168.26.100:2379"
192.168.26.102 | CHANGED | rc=0 >>
apiVersion: v1
kind: calicoApiConfig
metadata:
spec:
  datastoreType: "etcdv2"
  etcdEndpoints: "http://192.168.26.102:2379"
192.168.26.101 | CHANGED | rc=0 >>
apiVersion: v1
kind: calicoApiConfig
metadata:
spec:
  datastoreType: "etcdv2"
  etcdEndpoints: "http://192.168.26.101:2379"

实验相关镜像导入

┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$ansible etcd -m copy -a "src=/root/calico-node-v2.tar dest=/root/"
┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$ansible etcd -m shell -a "docker load -i /root/calico-node-v2.tar"
192.168.26.100 | CHANGED | rc=0 >>
Loaded image: quay.io/calico/node:v2.6.12
192.168.26.102 | CHANGED | rc=0 >>
Loaded image: quay.io/calico/node:v2.6.12
192.168.26.101 | CHANGED | rc=0 >>
Loaded image: quay.io/calico/node:v2.6.12
┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$

镜像查看

┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$ansible etcd -m shell -a "docker images"
192.168.26.102 | CHANGED | rc=0 >>
REPOSITORY            TAG       IMAGE ID       CREATED       SIZE
quay.io/calico/node   v2.6.12   401cc3e56a1a   3 years ago   281MB
192.168.26.100 | CHANGED | rc=0 >>
REPOSITORY            TAG       IMAGE ID       CREATED       SIZE
quay.io/calico/node   v2.6.12   401cc3e56a1a   3 years ago   281MB
192.168.26.101 | CHANGED | rc=0 >>
REPOSITORY            TAG       IMAGE ID       CREATED       SIZE
quay.io/calico/node   v2.6.12   401cc3e56a1a   3 years ago   281MB
┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$

calicoctl 工具导入

1 2	┌──[root@vms81.liruilongs.github.io]-[~/ansible] └─$ansible etcd -m copy -a "src=/root/calicoctl dest=/bin/ mode=+x"

搭建Calico网络

开始建立 calico node 信息：每个主机上都部署了Calico/Node作为虚拟路由器

1 2	┌──[root@vms81.liruilongs.github.io]-[~/ansible] └─$ansible etcd -m shell -a "calicoctl node run --node-image=quay.io/calico/node:v2.6.12 -c /etc/calico/calicoctl.cfg"

查看node状态，通过Calico将宿主机组织成任意的拓扑集群

┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$ansible etcd -m shell -a "calicoctl node status"
192.168.26.102 | CHANGED | rc=0 >>
Calico process is running.

IPv4 BGP status
+----------------+-------------------+-------+----------+-------------+
|  PEER ADDRESS  |     PEER TYPE     | STATE |  SINCE   |    INFO     |
+----------------+-------------------+-------+----------+-------------+
| 192.168.26.100 | node-to-node mesh | up    | 14:46:35 | Established |
| 192.168.26.101 | node-to-node mesh | up    | 14:46:34 | Established |
+----------------+-------------------+-------+----------+-------------+

IPv6 BGP status
No IPv6 peers found.
192.168.26.101 | CHANGED | rc=0 >>
Calico process is running.

IPv4 BGP status
+----------------+-------------------+-------+----------+-------------+
|  PEER ADDRESS  |     PEER TYPE     | STATE |  SINCE   |    INFO     |
+----------------+-------------------+-------+----------+-------------+
| 192.168.26.100 | node-to-node mesh | up    | 14:46:31 | Established |
| 192.168.26.102 | node-to-node mesh | up    | 14:46:34 | Established |
+----------------+-------------------+-------+----------+-------------+

IPv6 BGP status
No IPv6 peers found.
192.168.26.100 | CHANGED | rc=0 >>
Calico process is running.

IPv4 BGP status
+----------------+-------------------+-------+----------+-------------+
|  PEER ADDRESS  |     PEER TYPE     | STATE |  SINCE   |    INFO     |
+----------------+-------------------+-------+----------+-------------+
| 192.168.26.101 | node-to-node mesh | up    | 14:46:31 | Established |
| 192.168.26.102 | node-to-node mesh | up    | 14:46:35 | Established |
+----------------+-------------------+-------+----------+-------------+

IPv6 BGP status
No IPv6 peers found.
┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$

当集群中的容器需要与外界通信时,就可以通过BGP协议将网关物理路由器加入到集群中,使外界可以直接访问容器IP,而不需要做任何NAT之类的复杂操作。

通过Calico网络实现跨主机通信

在某一个Node上创建一个docker内部calico网络

┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$ansible 192.168.26.100 -m shell -a "docker network create --driver calico --ipam-driver calico-ipam calnet1"
192.168.26.100 | CHANGED | rc=0 >>
58121f89bcddec441770aa207ef662d09e4413625b0827ce4d8f601fb10650d0

会发现这个内网网络变成的一个全局的网络，在所有节点可见，58121f89bcdd

┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$ansible etcd -m shell -a "docker network list"
192.168.26.100 | CHANGED | rc=0 >>
NETWORK ID     NAME      DRIVER    SCOPE
caa87ba3dd86   bridge    bridge    local
58121f89bcdd   calnet1   calico    global
1d63e3ad385f   host      host      local
adc94f172d5f   none      null      local
192.168.26.102 | CHANGED | rc=0 >>
NETWORK ID     NAME      DRIVER    SCOPE
cc37d3c66e2f   bridge    bridge    local
58121f89bcdd   calnet1   calico    global
3b138015d4ab   host      host      local
7481614a7084   none      null      local
192.168.26.101 | CHANGED | rc=0 >>
NETWORK ID     NAME      DRIVER    SCOPE
d0cb224ed111   bridge    bridge    local
58121f89bcdd   calnet1   calico    global
106e1c9fb3d3   host      host      local
f983021e2a02   none      null      local
┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$

查看节点中的网卡信息,这个时候没有容器运行，所以没有caliao网卡

┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$ansible etcd -m shell -a "ip a"
192.168.26.102 | CHANGED | rc=0 >>
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN qlen 1
    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: ens32: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UP qlen 1000
    link/ether 00:0c:29:0f:98:f1 brd ff:ff:ff:ff:ff:ff
    inet 192.168.26.102/24 brd 192.168.26.255 scope global ens32
       valid_lft forever preferred_lft forever
    inet6 fe80::20c:29ff:fe0f:98f1/64 scope link
       valid_lft forever preferred_lft forever
3: docker0: <NO-CARRIER,BROADCAST,MULTICAST,UP> mtu 1500 qdisc noqueue state DOWN
    link/ether 02:42:c3:28:19:78 brd ff:ff:ff:ff:ff:ff
    inet 172.17.0.1/16 brd 172.17.255.255 scope global docker0
       valid_lft forever preferred_lft forever
192.168.26.100 | CHANGED | rc=0 >>
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN qlen 1
    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: ens32: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UP qlen 1000
    link/ether 00:0c:29:8c:e8:1a brd ff:ff:ff:ff:ff:ff
    inet 192.168.26.100/24 brd 192.168.26.255 scope global ens32
       valid_lft forever preferred_lft forever
    inet6 fe80::20c:29ff:fe8c:e81a/64 scope link
       valid_lft forever preferred_lft forever
3: docker0: <NO-CARRIER,BROADCAST,MULTICAST,UP> mtu 1500 qdisc noqueue state DOWN
    link/ether 02:42:f7:1a:2e:30 brd ff:ff:ff:ff:ff:ff
    inet 172.17.0.1/16 brd 172.17.255.255 scope global docker0
       valid_lft forever preferred_lft forever
192.168.26.101 | CHANGED | rc=0 >>
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN qlen 1
    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: ens32: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UP qlen 1000
    link/ether 00:0c:29:3b:6e:ef brd ff:ff:ff:ff:ff:ff
    inet 192.168.26.101/24 brd 192.168.26.255 scope global ens32
       valid_lft forever preferred_lft forever
    inet6 fe80::20c:29ff:fe3b:6eef/64 scope link
       valid_lft forever preferred_lft forever
3: docker0: <NO-CARRIER,BROADCAST,MULTICAST,UP> mtu 1500 qdisc noqueue state DOWN
    link/ether 02:42:70:a7:4e:7e brd ff:ff:ff:ff:ff:ff
    inet 172.17.0.1/16 brd 172.17.255.255 scope global docker0
       valid_lft forever preferred_lft forever
┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$

每个节点运行一个容器

┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$ansible etcd -m shell -a "docker run --name {{inventory_hostname}} -itd --net=calnet1 --restart=always busybox "
192.168.26.101 | CHANGED | rc=0 >>
cf2ff4b65e6343fa6e9afba6e75376b97ac47ea59c35f3c492bb7051c15627f0
192.168.26.100 | CHANGED | rc=0 >>
065724c073ded04d6df41d295be3cd5585f8683664fd42a3953dc8067195c58e
192.168.26.102 | CHANGED | rc=0 >>
82e4d6dfde5a6e51f9a4d4f86909678a42e8d1e2d9bfa6edd9cc258b37dfc2db

查看容器节点信息

┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$ansible etcd -m shell -a "docker ps"
192.168.26.102 | CHANGED | rc=0 >>
CONTAINER ID   IMAGE                         COMMAND         CREATED              STATUS              PORTS     NAMES
82e4d6dfde5a   busybox                       "sh"            About a minute ago   Up About a minute             192.168.26.102
c2d2ab904d6d   quay.io/calico/node:v2.6.12   "start_runit"   2 hours ago          Up 2 hours                    calico-node
192.168.26.100 | CHANGED | rc=0 >>
CONTAINER ID   IMAGE                         COMMAND         CREATED              STATUS              PORTS     NAMES
065724c073de   busybox                       "sh"            About a minute ago   Up About a minute             192.168.26.100
f0b150a924d9   quay.io/calico/node:v2.6.12   "start_runit"   2 hours ago          Up 2 hours                    calico-node
192.168.26.101 | CHANGED | rc=0 >>
CONTAINER ID   IMAGE                         COMMAND         CREATED              STATUS              PORTS     NAMES
cf2ff4b65e63   busybox                       "sh"            About a minute ago   Up About a minute             192.168.26.101
0e4e6f005797   quay.io/calico/node:v2.6.12   "start_runit"   2 hours ago          Up 2 hours                    calico-node
┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$

查看每个容器的内部网卡和IP

┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$ansible etcd -m shell -a "docker exec -it {{inventory_hostname}} ip a | grep  cali0 -A 4"
192.168.26.100 | CHANGED | rc=0 >>
4: cali0@if5: <BROADCAST,MULTICAST,UP,LOWER_UP,M-DOWN> mtu 1500 qdisc noqueue
    link/ether ee:ee:ee:ee:ee:ee brd ff:ff:ff:ff:ff:ff
    inet 192.168.239.128/32 scope global cali0
       valid_lft forever preferred_lft forever
192.168.26.102 | CHANGED | rc=0 >>
4: cali0@if5: <BROADCAST,MULTICAST,UP,LOWER_UP,M-DOWN> mtu 1500 qdisc noqueue
    link/ether ee:ee:ee:ee:ee:ee brd ff:ff:ff:ff:ff:ff
    inet 192.168.63.64/32 scope global cali0
       valid_lft forever preferred_lft forever
192.168.26.101 | CHANGED | rc=0 >>
4: cali0@if5: <BROADCAST,MULTICAST,UP,LOWER_UP,M-DOWN> mtu 1500 qdisc noqueue
    link/ether ee:ee:ee:ee:ee:ee brd ff:ff:ff:ff:ff:ff
    inet 192.168.198.0/32 scope global cali0
       valid_lft forever preferred_lft forever
┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$

查看容器内的路由关系，即所有的出口都是通过cali0网卡来实现的

┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$ansible etcd -m shell -a "docker exec -it {{inventory_hostname}} ip route | grep  cali0 "
192.168.26.101 | CHANGED | rc=0 >>
default via 169.254.1.1 dev cali0
169.254.1.1 dev cali0 scope link
192.168.26.102 | CHANGED | rc=0 >>
default via 169.254.1.1 dev cali0
169.254.1.1 dev cali0 scope link
192.168.26.100 | CHANGED | rc=0 >>
default via 169.254.1.1 dev cali0
169.254.1.1 dev cali0 scope link

每创建一个容器，则会在物理机上创建一张虚拟网卡出来，对应容器中的网卡，从这里可以看到容器里的虚拟网卡 cali0 和物理机的 cali6f956c2ada9 是 veth pair 关系。

关于veth pair 小伙伴可以百度下，这里简单描述，作用很简单，就是要把从一个 network namespace 发出的数据包转发到另一个 namespace。veth 设备是成对的，一个是container之中，另一个在container之外(宿主机)，即在真实机器上能看到的。VETH设备总是成对出现，送到一端请求发送的数据总是从另一端以请求接受的形式出现。创建并配置正确后，向其一端输入数据，VETH会改变数据的方向并将其送入内核网络子系统，完成数据的注入，而在另一端则能读到此数据。(Namespace，其中往veth设备上任意一端上RX到的数据，都会在另一端上以TX的方式发送出去)veth工作在L2数据链路层，veth-pair设备在转发数据包过程中并不串改数据包内容。

更多小伙伴可以参考:https://blog.csdn.net/sld880311/article/details/77650937

┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$ansible etcd -m shell -a "ip a | grep -A 4 cali"
192.168.26.102 | CHANGED | rc=0 >>
5: cali6f956c2ada9@if4: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP
    link/ether 6a:65:54:1a:19:e6 brd ff:ff:ff:ff:ff:ff link-netnsid 0
    inet6 fe80::6865:54ff:fe1a:19e6/64 scope link
       valid_lft forever preferred_lft forever
192.168.26.100 | CHANGED | rc=0 >>
5: cali0b7f49da20a@if4: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP
    link/ether 9e:da:0e:cc:b3:7e brd ff:ff:ff:ff:ff:ff link-netnsid 0
    inet6 fe80::9cda:eff:fecc:b37e/64 scope link
       valid_lft forever preferred_lft forever
192.168.26.101 | CHANGED | rc=0 >>
5: calib6f7ddae7e3@if4: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP
    link/ether 1e:e6:16:ae:f0:91 brd ff:ff:ff:ff:ff:ff link-netnsid 0
    inet6 fe80::1ce6:16ff:feae:f091/64 scope link
       valid_lft forever preferred_lft forever
┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$

查看宿主机路由关系

┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$ansible etcd -m shell -a "ip route "
192.168.26.101 | CHANGED | rc=0 >>
default via 192.168.26.2 dev ens32
169.254.0.0/16 dev ens32 scope link metric 1002
172.17.0.0/16 dev docker0 proto kernel scope link src 172.17.0.1
192.168.26.0/24 dev ens32 proto kernel scope link src 192.168.26.101
192.168.63.64/26 via 192.168.26.102 dev ens32 proto bird
blackhole 192.168.198.0/26 proto bird
192.168.198.1 dev cali2f9e2c68bad scope link
192.168.239.128/26 via 192.168.26.100 dev ens32 proto bird
192.168.26.100 | CHANGED | rc=0 >>
default via 192.168.26.2 dev ens32
169.254.0.0/16 dev ens32 scope link metric 1002
172.17.0.0/16 dev docker0 proto kernel scope link src 172.17.0.1
192.168.26.0/24 dev ens32 proto kernel scope link src 192.168.26.100
192.168.63.64/26 via 192.168.26.102 dev ens32 proto bird
192.168.198.0/26 via 192.168.26.101 dev ens32 proto bird
192.168.239.128 dev cali0b7f49da20a scope link
blackhole 192.168.239.128/26 proto bird
192.168.26.102 | CHANGED | rc=0 >>
default via 192.168.26.2 dev ens32
169.254.0.0/16 dev ens32 scope link metric 1002
172.17.0.0/16 dev docker0 proto kernel scope link src 172.17.0.1
192.168.26.0/24 dev ens32 proto kernel scope link src 192.168.26.102
192.168.63.64 dev cali6f956c2ada9 scope link
blackhole 192.168.63.64/26 proto bird
192.168.198.0/26 via 192.168.26.101 dev ens32 proto bird
192.168.239.128/26 via 192.168.26.100 dev ens32 proto bird
┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$

我们那其中一台机器来看：192.168.26.100宿主机来讲

192.168.239.128 dev cali0b7f49da20a scope link

进去：本机到目的地址到容器IP(192.168.239.128 ) 的数据包都从 cali6f956c2ada9 (新产生的虚拟网卡)走。

192.168.63.64/26 via 192.168.26.102 dev ens32 proto bird
192.168.198.0/26 via 192.168.26.101 dev ens32 proto bird

出来：本机目的地址到容器IP(192.168.63.64/26) 容器IP(192.168.198.0/26) 网段的数据包都从 ens32 发到其他的两个宿主机上去。

每台主机都知道不同的容器在哪台主机上，所以会动态的设置路由。

┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$ansible etcd -m shell -a "route -n "
192.168.26.101 | CHANGED | rc=0 >>
Kernel IP routing table
Destination     Gateway         Genmask         Flags Metric Ref    Use Iface
0.0.0.0         192.168.26.2    0.0.0.0         UG    0      0        0 ens32
169.254.0.0     0.0.0.0         255.255.0.0     U     1002   0        0 ens32
172.17.0.0      0.0.0.0         255.255.0.0     U     0      0        0 docker0
192.168.26.0    0.0.0.0         255.255.255.0   U     0      0        0 ens32
192.168.63.64   192.168.26.102  255.255.255.192 UG    0      0        0 ens32
192.168.198.0   0.0.0.0         255.255.255.192 U     0      0        0 *
192.168.198.1   0.0.0.0         255.255.255.255 UH    0      0        0 cali2f9e2c68bad
192.168.239.128 192.168.26.100  255.255.255.192 UG    0      0        0 ens32
192.168.26.100 | CHANGED | rc=0 >>
Kernel IP routing table
Destination     Gateway         Genmask         Flags Metric Ref    Use Iface
0.0.0.0         192.168.26.2    0.0.0.0         UG    0      0        0 ens32
169.254.0.0     0.0.0.0         255.255.0.0     U     1002   0        0 ens32
172.17.0.0      0.0.0.0         255.255.0.0     U     0      0        0 docker0
192.168.26.0    0.0.0.0         255.255.255.0   U     0      0        0 ens32
192.168.63.64   192.168.26.102  255.255.255.192 UG    0      0        0 ens32
192.168.198.0   192.168.26.101  255.255.255.192 UG    0      0        0 ens32
192.168.239.128 0.0.0.0         255.255.255.255 UH    0      0        0 cali0b7f49da20a
192.168.239.128 0.0.0.0         255.255.255.192 U     0      0        0 *
192.168.26.102 | CHANGED | rc=0 >>
Kernel IP routing table
Destination     Gateway         Genmask         Flags Metric Ref    Use Iface
0.0.0.0         192.168.26.2    0.0.0.0         UG    0      0        0 ens32
169.254.0.0     0.0.0.0         255.255.0.0     U     1002   0        0 ens32
172.17.0.0      0.0.0.0         255.255.0.0     U     0      0        0 docker0
192.168.26.0    0.0.0.0         255.255.255.0   U     0      0        0 ens32
192.168.63.64   0.0.0.0         255.255.255.255 UH    0      0        0 cali6f956c2ada9
192.168.63.64   0.0.0.0         255.255.255.192 U     0      0        0 *
192.168.198.0   192.168.26.101  255.255.255.192 UG    0      0        0 ens32
192.168.239.128 192.168.26.100  255.255.255.192 UG    0      0        0 ens32
┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$

简单测试一下：192.168.26.100宿主机上的容器(192.168.239.128)去ping 192.168.63.64(192.168.26.100上的容器)，实现跨主机互通。

┌──[root@vms100.liruilongs.github.io]-[~]
└─$ docker exec -it 192.168.26.100 /bin/sh
/ # ls
bin   dev   etc   home  proc  root  sys   tmp   usr   var
/ # ip a
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue qlen 1
    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
4: cali0@if5: <BROADCAST,MULTICAST,UP,LOWER_UP,M-DOWN> mtu 1500 qdisc noqueue
    link/ether ee:ee:ee:ee:ee:ee brd ff:ff:ff:ff:ff:ff
    inet 192.168.239.128/32 scope global cali0
       valid_lft forever preferred_lft forever
/ # ping 192.168.63.64
PING 192.168.63.64 (192.168.63.64): 56 data bytes
64 bytes from 192.168.63.64: seq=0 ttl=62 time=18.519 ms
64 bytes from 192.168.63.64: seq=1 ttl=62 time=0.950 ms
64 bytes from 192.168.63.64: seq=2 ttl=62 time=1.086 ms
64 bytes from 192.168.63.64: seq=3 ttl=62 time=0.846 ms
64 bytes from 192.168.63.64: seq=4 ttl=62 time=0.840 ms
64 bytes from 192.168.63.64: seq=5 ttl=62 time=1.151 ms
64 bytes from 192.168.63.64: seq=6 ttl=62 time=0.888 ms
^C
--- 192.168.63.64 ping statistics ---
7 packets transmitted, 7 packets received, 0% packet loss
round-trip min/avg/max = 0.840/3.468/18.519 ms
/ #

在K8s集群的中，有一个容器，就会生成一个calico网卡

┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$ip a
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN qlen 1
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
    inet 127.0.0.1/8 scope host lo
       valid_lft forever preferred_lft forever
    inet6 ::1/128 scope host
       valid_lft forever preferred_lft forever
2: ens32: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UP qlen 1000
    link/ether 00:0c:29:ad:e3:93 brd ff:ff:ff:ff:ff:ff
    inet 192.168.26.81/24 brd 192.168.26.255 scope global ens32
       valid_lft forever preferred_lft forever
    inet6 fe80::20c:29ff:fead:e393/64 scope link
       valid_lft forever preferred_lft forever
3: docker0: <NO-CARRIER,BROADCAST,MULTICAST,UP> mtu 1500 qdisc noqueue state DOWN
    link/ether 02:42:0a:9e:7d:44 brd ff:ff:ff:ff:ff:ff
    inet 172.17.0.1/16 brd 172.17.255.255 scope global docker0
       valid_lft forever preferred_lft forever
4: tunl0@NONE: <NOARP,UP,LOWER_UP> mtu 1480 qdisc noqueue state UNKNOWN qlen 1
    link/ipip 0.0.0.0 brd 0.0.0.0
    inet 10.244.88.64/32 scope global tunl0
       valid_lft forever preferred_lft forever
5: cali12cf25006b5@if4: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1480 qdisc noqueue state UP
    link/ether ee:ee:ee:ee:ee:ee brd ff:ff:ff:ff:ff:ff link-netnsid 0
    inet6 fe80::ecee:eeff:feee:eeee/64 scope link
       valid_lft forever preferred_lft forever
6: cali5a282a7bbb0@if4: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1480 qdisc noqueue state UP
    link/ether ee:ee:ee:ee:ee:ee brd ff:ff:ff:ff:ff:ff link-netnsid 1
    inet6 fe80::ecee:eeff:feee:eeee/64 scope link
       valid_lft forever preferred_lft forever
7: calicb34164ec79@if4: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1480 qdisc noqueue state UP
    link/ether ee:ee:ee:ee:ee:ee brd ff:ff:ff:ff:ff:ff link-netnsid 2
    inet6 fe80::ecee:eeff:feee:eeee/64 scope link
       valid_lft forever preferred_lft forever
┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$

Kubernetes网络策略

为了实现细粒度的容器间网络访问隔离策略(防火墙), Kubernetes从1.3版本开始,由SIG-Network小组主导研发了Network Policy机制,目前已升级为networking.k8s.io/v1稳定版本。

Network Policy的主要功能是对Pod间的网络通信进行限制和准入控制

设置方式为将Pod的Label作为查询条件,设置允许访问或禁止访问的客户端Pod列表。查询条件可以作用于Pod和Namespace级别。

为了使用Network Policy, Kubernetes引入了一个新的资源对象NetworkPolicy,供用户设置Pod间网络访问的策略。但仅定义一个网络策略是无法完成实际的网络隔离的,还需要一个策略控制器(PolicyController)进行策略的实现。

策略控制器由第三方网络组件提供,目前Calico, Cilium, Kube-router, Romana, Weave Net等开源项目均支持网络策略的实现。Network Policy的工作原理如图

在这里插入图片描述

policy controller需要实现一个API Listener,监听用户设置的NetworkPolicy定义,并将网络访问规则通过各Node的Agent进行实际设置(Agent则需要通过CNI网络插件实现)

网络策略配置说明

网络策略的设置主要用于对目标Pod的网络访问进行限制,在默认·情况下对所有Pod都是允许访问的,在设置了指向Pod的NetworkPolicy网络策略之后,到Pod的访问才会被限制。 需要注意的是网络策略是基于Pod的

NetWorkPolicy基于命名空间进行限制，即只作用当前命名空间，分为两种：

ingress:定义允许访问目标Pod的入站白名单规则
egress: 定义目标Pod允许访问的“出站”白名单规则

具体的规则限制方式分为三种(需要注意的是，多个限制之间是或的逻辑关系,如果希望变成与的关系，yaml文件需要配置为数组)：

IP 策略
命名空间策略
pod选择器限制

下面是一个资源文件的Demo

apiVersion: networking.k8s.io/v1
kind: NetworkPolicy
metadata:
  name: test-network-policy
  namespace: default
spec:
  podSelector: #用于定义该网络策略作用的Pod范围
    matchLabels:
      role: db
  policyTypes: #网络策略的类型，包括ingress和egress两种
  - Ingress
  - Egress
  ingress: #定义允许访问目标Pod的入站白名单规则
  - from: #满足from 条件的客户端才能访问ports定义的目标Pod端口号。
    - ipBlock: # IP限制
        cidr: 172.17.0.0/16
        except:  #排除那些IP
        - 172.17.1.0/24
    - namespaceSelector:  #命名空间限制
        matchLabels:
          project: myproject
    - podSelector:  # pod选择器限制
        matchLabels:
          role: frontend
    ports: #允许访问的目标Pod监听的端口号。
    - protocol: TCP
      port: 6379
  egress: #定义目标Pod允许访问的“出站”白名单规则
  - to: #目标Pod仅允许访问满足to条件的服务端IP范围和ports定义的端口号
    - ipBlock:
        cidr: 10.0.0.0/24
    ports:  #允许访问的服务端的端口号。
    - protocol: TCP
      port: 5978

在Namespace级别设置默认的网络策略

在Namespace级别还可以设置一些默认的全局网络策略,以方便管理员对整个Namespace进行统一的网络策略设置。

默认拒绝所有入站流量

apiVersion: networking.k8s.io/v1
kind: NetworkPolicy
metadata:
  name: default-deny-ingress
spec:
  podSelector: {}
  policyTypes:
  - Ingress

默认允许所有入站流量

apiVersion: networking.k8s.io/v1
kind: NetworkPolicy
metadata:
  name: allow-all-ingress
spec:
  podSelector: {}
  ingress:
  - {}
  policyTypes:
  - Ingress

默认拒绝所有出站流量

apiVersion: networking.k8s.io/v1
kind: NetworkPolicy
metadata:
  name: default-deny-egress
spec:
  podSelector: {}
  policyTypes:
  - Egress

默认允许所有出站流量

apiVersion: networking.k8s.io/v1
kind: NetworkPolicy
metadata:
  name: allow-all-egress
spec:
  podSelector: {}
  egress:
  - {}
  policyTypes:
  - Egress

默认拒绝所有入口和所有出站流量

apiVersion: networking.k8s.io/v1
kind: NetworkPolicy
metadata:
  name: default-deny-all
spec:
  podSelector: {}
  policyTypes:
  - Ingress
  - Egress

NetworkPolicy的发展

作为一个稳定特性，SCTP 支持默认是被启用的。要在集群层面禁用 SCTP，你(或你的集群管理员)需要为API服务器指定 --feature-gates=SCTPSupport=false,… 来禁用 SCTPSupport 特性门控。启用该特性门控后，用户可以将 NetworkPolicy 的protocol字段设置为 SCTP(不同版本略有区别)

NetWorkPolicy实战

环境准备

先创建两个没有任何策略的SVC

┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$d=k8s-network-create
┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$mkdir $d;cd $d
┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-network-create]
└─$kubectl  create  ns liruilong-network-create
namespace/liruilong-network-create created
┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-network-create]
└─$kubectl  config  set-context  $(kubectl config current-context) --namespace=liruilong-network-createContext "kubernetes-admin@kubernetes" modified.
┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-network-create]
└─$kubectl  config view | grep namespace
    namespace: liruilong-network-create
┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-network-create]
└─$

我们先构造两个pod,为两个SVC提供能力

┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-network-create]
└─$kubectl  run pod1 --image=nginx --image-pull-policy=IfNotPresent
pod/pod1 created
┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-network-create]
└─$kubectl  run pod2 --image=nginx --image-pull-policy=IfNotPresent
pod/pod2 created
┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-network-create]
└─$kubectl  get pods -o wide
NAME   READY   STATUS    RESTARTS   AGE   IP               NODE                         NOMINATED NODE   READINESS GATES
pod1   1/1     Running   0          35s   10.244.70.31     vms83.liruilongs.github.io   <none>           <none>
pod2   1/1     Running   0          21s   10.244.171.181   vms82.liruilongs.github.io   <none>           <none>
┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-network-create]
└─$

然后我们分别修改pod中Ngixn容器主页

┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-network-create]
└─$kubectl  get pods  --show-labels
NAME   READY   STATUS    RESTARTS   AGE    LABELS
pod1   1/1     Running   0          100s   run=pod1
pod2   1/1     Running   0          86s    run=pod2
┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-network-create]
└─$kubectl  exec  -it pod1 -- sh -c "echo pod1 >/usr/share/nginx/html/index.html"
┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-network-create]
└─$kubectl  exec  -it pod2 -- sh -c "echo pod2 >/usr/share/nginx/html/index.html"

创建两个SVC

┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-network-create]
└─$kubectl  expose  --name=svc1 pod pod1 --port=80 --type=LoadBalancer
service/svc1 exposed
┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-network-create]
└─$kubectl  expose  --name=svc2 pod pod2 --port=80 --type=LoadBalancer
service/svc2 exposed
┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-network-create]
└─$kubectl  get svc
NAME   TYPE           CLUSTER-IP       EXTERNAL-IP      PORT(S)        AGE
svc1   LoadBalancer   10.106.61.84     192.168.26.240   80:30735/TCP   14s
svc2   LoadBalancer   10.111.123.194   192.168.26.241   80:31034/TCP   5s
┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-network-create]
└─$

访问测试，无论在当前命名空间还是在指定命名空间，都可以相互访问

┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-network-create]
└─$kubectl  run testpod1 -it --rm --image=yauritux/busybox-curl --image-pull-policy=IfNotPresent
If you don''t see a command prompt, try pressing enter.
/home # curl svc1
pod1
/home # curl svc2
pod2
/home # exit

指定命名空间

┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-network-create]
└─$kubectl  run testpod2 -it --rm --image=yauritux/busybox-curl --image-pull-policy=IfNotPresent -n default
If you don''t see a command prompt, try pressing enter.
/home # curl svc1.liruilong-network-create
pod1
/home # curl svc2.liruilong-network-create
pod2
/home #

由于使用了LB，所以物理机也可以访问

PS E:\docker> curl 192.168.26.240   

StatusCode        : 200
StatusDescription : OK
Content           : pod1

RawContent        : HTTP/1.1 200 OK
                    Connection: keep-alive
                    Accept-Ranges: bytes
                    Content-Length: 5
                    Content-Type: text/html
                    Date: Mon, 03 Jan 2022 12:29:32 GMT
                    ETag: "61d27744-5"
                    Last-Modified: Mon, 03 Jan 2022 04:1...
Forms             : {}
Headers           : {[Connection, keep-alive], [Accept-Ranges, bytes], [Content-Lengt
                    h, 5], [Content-Type, text/html]...}
Images            : {}
InputFields       : {}
Links             : {}
ParsedHtml        : System.__ComObject
RawContentLength  : 5

PS E:\docker> curl 192.168.26.241

StatusCode        : 200
StatusDescription : OK
Content           : pod2

RawContent        : HTTP/1.1 200 OK
                    Connection: keep-alive
                    Accept-Ranges: bytes
                    Content-Length: 5
                    Content-Type: text/html
                    Date: Mon, 03 Jan 2022 12:29:49 GMT
                    ETag: "61d27752-5"
                    Last-Modified: Mon, 03 Jan 2022 04:1...
Forms             : {}
Headers           : {[Connection, keep-alive], [Accept-Ranges, bytes], [Content-Lengt
                    h, 5], [Content-Type, text/html]...}
Images            : {}
InputFields       : {}
Links             : {}
ParsedHtml        : System.__ComObject
RawContentLength  : 5
PS E:\docker>

进入策略

下面我们看一下进入的策略

┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-network-create]
└─$kubectl  get pods --show-labels
NAME   READY   STATUS    RESTARTS        AGE    LABELS
pod1   1/1     Running   2 (3d12h ago)   5d9h   run=pod1
pod2   1/1     Running   2 (3d12h ago)   5d9h   run=pod2
┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-network-create]
└─$kubectl  get svc -o wide
NAME   TYPE           CLUSTER-IP       EXTERNAL-IP      PORT(S)        AGE    SELECTOR
svc1   LoadBalancer   10.106.61.84     192.168.26.240   80:30735/TCP   5d9h   run=pod1
svc2   LoadBalancer   10.111.123.194   192.168.26.241   80:31034/TCP   5d9h   run=pod2
┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-network-create]
└─$

测试的外部物理机机器IP

PS E:\docker> ipconfig

Windows IP 配置
..........

以太网适配器 VMware Network Adapter VMnet8:

   连接特定的 DNS 后缀 . . . . . . . :
   本地链接 IPv6 地址. . . . . . . . : fe80::f9c8:e941:4deb:698f%24
   IPv4 地址 . . . . . . . . . . . . : 192.168.26.1
   子网掩码  . . . . . . . . . . . . : 255.255.255.0
   默认网关. . . . . . . . . . . . . :

IP限制

我们通过修改ip限制来演示网路策略，通过宿主机所在物理机访问。当设置指定网段可以访问，不是指定网段不可以访问

┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-network-create]
└─$vim networkpolicy.yaml
┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-network-create]
└─$kubectl  apply -f networkpolicy.yaml
networkpolicy.networking.k8s.io/test-network-policy configured

编写资源文件，允许172.17.0.0/16网段的机器访问

apiVersion: networking.k8s.io/v1
kind: NetworkPolicy
metadata:
  name: test-network-policy
  namespace: liruilong-network-create
spec:
  podSelector:
    matchLabels:
      run: pod1
  policyTypes:
  - Ingress
  ingress:
  - from:
    - ipBlock:
        cidr: 172.17.0.0/16 #  只允许这个网段访问
    ports:
    - protocol: TCP
      port: 80

集群外部机器无法访问

PS E:\docker> curl 192.168.26.240
curl : 无法连接到远程服务器
所在位置 行:1 字符: 1
+ curl 192.168.26.240
+ ~~~~~~~~~~~~~~~~~~~
    + CategoryInfo          : InvalidOperation: (System.Net.HttpWebRequest:HttpWebRequest) [Invoke-WebRequest]，WebExce
    ption
    + FullyQualifiedErrorId : WebCmdletWebResponseException,Microsoft.PowerShell.Commands.InvokeWebRequestCommand

配置允许当前网段的ip访问

apiVersion: networking.k8s.io/v1
kind: NetworkPolicy
metadata:
  name: test-network-policy
  namespace: liruilong-network-create
spec:
  podSelector:
    matchLabels:
      run: pod1
  policyTypes:
  - Ingress
  ingress:
  - from:
    - ipBlock:
        cidr: 192.168.26.0/24 #  只允许这个网段访问
    ports:
    - protocol: TCP
      port: 80

修改网段之后正常外部机器可以访问

┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-network-create]
└─$sed -i 's#172.17.0.0/16#192.168.26.0/24#' networkpolicy.yaml
┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-network-create]
└─$kubectl  apply -f networkpolicy.yaml

测试，外部机器可以访问

PS E:\docker> curl 192.168.26.240


StatusCode        : 200
StatusDescription : OK
Content           : pod1

RawContent        : HTTP/1.1 200 OK
                    Connection: keep-alive
                    Accept-Ranges: bytes
                    Content-Length: 5
                    Content-Type: text/html
                    Date: Sat, 08 Jan 2022 14:59:13 GMT
                    ETag: "61d9a663-5"
                    Last-Modified: Sat, 08 Jan 2022 14:5...
Forms             : {}
Headers           : {[Connection, keep-alive], [Accept-Ranges, bytes], [Content-Length, 5], [Content-T
                    ype, text/html]...}
Images            : {}
InputFields       : {}
Links             : {}
ParsedHtml        : System.__ComObject
RawContentLength  : 5

命名空间限制

设置只允许default命名空间的数据通过

┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-network-create]
└─$kubectl  get ns --show-labels  | grep default
default                      Active   26d     kubernetes.io/metadata.name=default
┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-network-create]
└─$vim networkpolicy-name.yaml
┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-network-create]
└─$kubectl apply -f  networkpolicy-name.yaml
networkpolicy.networking.k8s.io/test-network-policy configured

apiVersion: networking.k8s.io/v1
kind: NetworkPolicy
metadata:
  name: test-network-policy
  namespace: liruilong-network-create
spec:
  podSelector:
    matchLabels:
      run: pod1
  policyTypes:
  - Ingress
  ingress:
  - from:
    - namespaceSelector:
        matchLabels:
          kubernetes.io/metadata.name: default
    ports:
    - protocol: TCP
      port: 80

宿主机所在物理机无法访问访问

PS E:\docker> curl 192.168.26.240
curl : 无法连接到远程服务器
所在位置 行:1 字符: 1
+ curl 192.168.26.240
+ ~~~~~~~~~~~~~~~~~~~
    + CategoryInfo          : InvalidOperation: (System.Net.HttpWebRequest:HttpWebRequest) [Invoke-WebR
   equest]，WebException
    + FullyQualifiedErrorId : WebCmdletWebResponseException,Microsoft.PowerShell.Commands.InvokeWebRequ
   estCommand

PS E:\docker>

当前命名空间也无法访问

┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-network-create]
└─$kubectl  run testpod1 -it --rm --image=yauritux/busybox-curl --image-pull-policy=IfNotPresent

/home # curl --connect-timeout 10 -m 10 svc1
curl: (28) Connection timed out after 10413 milliseconds

default命名空间可以访问

┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-network-create]
└─$kubectl  run testpod1 -it --rm --image=yauritux/busybox-curl --image-pull-policy=IfNotPresent --namespace=default

/home # curl  --connect-timeout 10 -m 10 svc1.liruilong-network-create
pod1
/home #

pod选择器限制

apiVersion: networking.k8s.io/v1
kind: NetworkPolicy
metadata:
  name: test-network-policy
  namespace: liruilong-network-create
spec:
  podSelector:
    matchLabels:
      run: pod1
  policyTypes:
  - Ingress
  ingress:
  - from:
    - podSelector:
        matchLabels:
          run: testpod
    ports:
    - protocol: TCP
      port: 80

创建一个策略，只允许标签为run=testpod的pod访问

1
2
3

┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-network-create]
└─$kubectl  apply  -f networkpolicy-pod.yaml
networkpolicy.networking.k8s.io/test-network-policy created

创建两个pod，都设置--labels=run=testpod 标签，只有当前命名空间可以访问

┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-network-create]
└─$kubectl  run testpod1 -it --rm --image=yauritux/busybox-curl --image-pull-policy=IfNotPresent --labels=run=testpod --namespace=default

/home # curl  --connect-timeout 10 -m 10 svc1.liruilong-network-create
curl: (28) Connection timed out after 10697 milliseconds

┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-network-create]
└─$kubectl  run testpod1 -it --rm --image=yauritux/busybox-curl --image-pull-policy=IfNotPresent --labels=run=testpod

/home # curl  --connect-timeout 10 -m 10 svc1
pod1
/home #

下面的设置可以其他所有命名空间可以访问

apiVersion: networking.k8s.io/v1
kind: NetworkPolicy
metadata:
  name: test-network-policy
  namespace: liruilong-network-create
spec:
  podSelector:
    matchLabels:
      run: pod1
  policyTypes:
  - Ingress
  ingress:
  - from:
    - podSelector:
        matchLabels:
      podSelector:
        matchLabels:
          run: testpod
    ports:
    - protocol: TCP
      port: 80

default 命名空间和当前命名空间可以访问

apiVersion: networking.k8s.io/v1
kind: NetworkPolicy
metadata:
  name: test-network-policy
  namespace: liruilong-network-create
spec:
  podSelector:
    matchLabels:
      run: pod1
  policyTypes:
  - Ingress
  ingress:
  - from:
    - podSelector:
        matchLabels:
          kubernetes.io/metadata.name: default
      podSelector:
        matchLabels:
          run: testpod
    - podSelector:
        matchLabels:
          run: testpod      
    ports:
    - protocol: TCP
      port: 80

定位pod所使用的网络策略

┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-network-create]
└─$kubectl  get networkpolicies
NAME                  POD-SELECTOR   AGE
test-network-policy   run=pod1       13m
┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-network-create]
└─$kubectl  get pods --show-labels
NAME   READY   STATUS    RESTARTS        AGE     LABELS
pod1   1/1     Running   2 (3d15h ago)   5d12h   run=pod1
pod2   1/1     Running   2 (3d15h ago)   5d12h   run=pod2
┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-network-create]
└─$kubectl  get networkpolicies | grep run=pod1
test-network-policy   run=pod1       15m
┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-network-create]
└─$

出去策略

apiVersion: networking.k8s.io/v1
kind: NetworkPolicy
metadata:
  name: test-network-policy
spec:
  podSelector:
    matchLabels:
      run: pod1
  policyTypes:
  - Egress
  egress:
  - to:
    - podSelector:
        matchLabels:
          run: pod2
    ports:
    - protocol: TCP
      port: 80

pod1只能访问pod2的TCP协议80端口

1 2	┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-network-create] └─$kubectl apply -f networkpolicy1.yaml

通过IP访问正常pod2

┌──[root@vms81.liruilongs.github.io]-[~]
└─$kubectl  exec -it pod1 -- bash
root@pod1:/# curl 10.111.123.194
pod2

因为DNS的pod在另一个命名空间(kube-system)运行，pod1只能到pod，所以无法通过域名访问，需要添加另一个命名空间

┌──[root@vms81.liruilongs.github.io]-[~]
└─$kubectl  exec -it pod1 -- bash
root@pod1:/# curl svc2
^C

相关参数获取

┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-network-create]
└─$kubectl  get ns --show-labels | grep kube-system
kube-system                  Active   27d     kubernetes.io/metadata.name=kube-system
┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-network-create]
└─$kubectl get pods  --show-labels -n kube-system  | grep dns
coredns-7f6cbbb7b8-ncd2s                             1/1     Running   13 (3d19h ago)    24d   k8s-app=kube-dns,pod-template-hash=7f6cbbb7b8
coredns-7f6cbbb7b8-pjnct                             1/1     Running   13 (3d19h ago)    24d   k8s-app=kube-dns,pod-template-hash=7f6cbbb7b8

配置两个出去规则，一个到pod2,一个到kube-dns，使用不同端口协议

apiVersion: networking.k8s.io/v1
kind: NetworkPolicy
metadata:
  name: test-network-policy
spec:
  podSelector:
    matchLabels:
      run: pod1
  policyTypes:
  - Egress
  egress:
  - to:
    - podSelector:
        matchLabels:
          run: pod2
    ports:
    - protocol: TCP
      port: 80
  - to:
    - namespaceSelector:
        matchLabels:
          kubernetes.io/metadata.name: kube-system
      podSelector:
        matchLabels:
          k8s-app: kube-dns
    ports:
    - protocol: UDP
      port: 53

测试可以通过域名访问

┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-network-create]
└─$vim networkpolicy2.yaml
┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-network-create]
└─$kubectl  apply  -f networkpolicy2.yaml
networkpolicy.networking.k8s.io/test-network-policy configured
┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-network-create]
└─$kubectl  get networkpolicies
NAME                  POD-SELECTOR   AGE
test-network-policy   run=pod1       3h38m
┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-network-create]
└─$kubectl  exec -it pod1 -- bash
root@pod1:/# curl svc2
pod2
root@pod1:/#

时间关系，关于网络策略和小伙伴们分享到这里，生活加油

关于 kubernetes中网络CNI方面的一些笔记

https://liruilongs.github.io/2021/12/28/K8s/API 学习/关于 kubernetes中网络方面的一些笔记/

作者

山河已无恙

发布于

2021-12-28

更新于

2023-06-21

关于 kubernetes中网络CNI方面的一些笔记

写在前面

Kubernetes网络

K8s中Calico的实现方案

跨主机Docker网络通信

calico通信过程

Calico实现方案

环境准备

搭建Calico网络

通过Calico网络实现跨主机通信

Kubernetes网络策略

网络策略配置说明

在Namespace级别设置默认的网络策略

默认拒绝所有入站流量

默认允许所有入站流量

默认拒绝所有出站流量

默认允许所有出站流量

默认拒绝所有入口和所有出站流量

NetworkPolicy的发展

NetWorkPolicy实战

环境准备

进入策略

IP限制

命名空间限制

pod选择器限制

定位pod所使用的网络策略

出去策略

作者

发布于

更新于

许可协议

喜欢这篇文章？打赏一下作者吧

目录

链接

最新评论

最新文章

分类

归档

标签

订阅更新

Your browser is out-of-date!