主要参考每天5分钟玩转docker容器技术博客。
记录下自己的实验过程。
如果想学习,请直接参考该博客http://www.cnblogs.com/CloudMan6/

网络概述

flannel为每个host分配一个subnet,容器从此subnet中分配ip。这些ip可以在host间路由,容器间无需nat和port mapping就可以跨主机通信。

每个subnet都是从一个更大的ip池中划分的,flannel会在每个主机上运行一个叫flanneld的agent,职责是从池中分配subnet。为了在各个主机间共享信息,flannel用etcd存放网络配置、已分配的subnet、host的ip等信息。

数据包如何在主机间转发是由 backend 实现的。flannel 提供了多种 backend,最常用的有 vxlan 和 host-gw

环境搭建

安装配置etcd

在192.168.19.120上运行如下脚本:

ETCD_VER=v2.3.7

DOWNLOAD_URL=https://github.com/coreos/etcd/releases/download

curl -L ${DOWNLOAD_URL}/${ETCD_VER}/etcd-${ETCD_VER}-linux-amd64.tar.gz -o /tmp/etcd-${ETCD_VER}-linux-amd64.tar.gz

mkdir -p /tmp/test-etcd && tar xzvf /tmp/etcd-${ETCD_VER}-linux-amd64.tar.gz -C /tmp/test-etcd --strip-components=1

cp /tmp/test-etcd/etcd* /usr/local/bin/

运行etcd

etcd -listen-client-urls http://192.168.19.120:2379 -advertise-client-urls http://192.168.19.120:2379

build flannel

  • 下载并重命名 image。

docker pull cloudman6/kube-cross:v1.6.2-2

docker tag cloudman6/kube-cross:v1.6.2-2 gcr.io/google_containers/kube-cross:v1.6.2-2

  • 下载 flannel 源码。

git clone https://github.com/coreos/fla...

  • 开始构建。

cd flannel
make dist/flanneld-amd64

  • 将flannel执行文件拷贝到host2 和host3

scp dist/flanneld-amd64 192.168.19.121:/usr/local/bin/flanneld
scp dist/flanneld-amd64 192.168.19.122:/usr/local/bin/flanneld

将flannel网络的配置文件保存到etcd中

配置文件如下:

{
  "Network":"10.2.0.0/16",
  "SubnetLen":24,
  "Backend":{
    "Type":"vxlan"
    }
}

解释:

  1. Network 定义该网络的ip池
  2. SubnetLen 指定每个主机分配到的subnet大小为24位,即 10.2.x.0/24
  3. Backend为vxlan,即主机间通过vxlan通信。

将配置存入etcd

etcdctl --endpoint=192.168.19.120:2379 set /docker-test/network/config < flannel_config.json

启动flannel

启动命令

在两台机器上运行如下命令:

flanneld -etcd-endpoints=http://192.168.19.120:2379 -iface=enp0s3 -etcd-prefix=/docker-test/network

-etcd-endpoints 指定 etcd url
-iface 指定主机间数据传输使用的interface
-etcd-prefix 指定etcd存放flannel网络配置文件的key

输出如下:

root@host3:~# flanneld -etcd-endpoints=http://192.168.19.120:2379 -iface=enp0s3 -etcd-prefix=/docker-test/network
I0520 02:24:57.594377    2463 main.go:499] Using interface with name enp0s3 and address 192.168.19.122
I0520 02:24:57.594766    2463 main.go:516] Defaulting external address to interface address (192.168.19.122)
I0520 02:24:57.595348    2463 main.go:240] Created subnet manager: Etcd Local Manager with Previous Subnet: None
I0520 02:24:57.595643    2463 main.go:243] Installing signal handlers
I0520 02:24:57.598847    2463 main.go:358] Found network config - Backend type: vxlan
I0520 02:24:57.599288    2463 vxlan.go:120] VXLAN config: VNI=1 Port=0 GBP=false DirectRouting=false
I0520 02:24:57.649392    2463 local_manager.go:234] Picking subnet in range 10.2.1.0 ... 10.2.255.0
I0520 02:24:57.651626    2463 local_manager.go:220] Allocated lease (10.2.99.0/24) to current node (192.168.19.122) 
I0520 02:24:57.653626    2463 main.go:305] Wrote subnet file to /run/flannel/subnet.env
I0520 02:24:57.654905    2463 main.go:309] Running backend.
I0520 02:24:57.660772    2463 vxlan_network.go:60] watching for new subnet leases
I0520 02:24:57.665879    2463 main.go:401] Waiting for 23h0m0.025221044s to renew lease
I0520 02:24:57.672707    2463 iptables.go:115] Some iptables rules are missing; deleting and recreating rules
I0520 02:24:57.674333    2463 iptables.go:137] Deleting iptables rule: -s 10.2.0.0/16 -j ACCEPT
I0520 02:24:57.678549    2463 iptables.go:137] Deleting iptables rule: -d 10.2.0.0/16 -j ACCEPT
I0520 02:24:57.688935    2463 iptables.go:125] Adding iptables rule: -s 10.2.0.0/16 -j ACCEPT
I0520 02:24:57.696815    2463 iptables.go:125] Adding iptables rule: -d 10.2.0.0/16 -j ACCEPT

启动后的效果

  • 新建了一个interface flannel.1, 且配置上subnet的第一个ip 10.2.21.0
root@host2:~# ip addr show flannel.1
8: flannel.1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1450 qdisc noqueue state UNKNOWN group default 
    link/ether 6e:9f:1d:bf:5c:b9 brd ff:ff:ff:ff:ff:ff
    inet 10.2.21.0/32 scope global flannel.1
       valid_lft forever preferred_lft forever
    inet6 fe80::6c9f:1dff:febf:5cb9/64 scope link 
       valid_lft forever preferred_lft forever
  • 添加了一条路由:目的地址为flannel网络10.2.0.0/16的数据包都有flannel.1转发
root@host2:~# ip r
default via 192.168.19.2 dev enp0s3 onlink 
10.2.99.0/24 via 10.2.99.0 dev flannel.1 onlink 
172.17.0.0/16 dev docker0  proto kernel  scope link  src 172.17.0.1 linkdown 
172.18.0.0/16 dev docker_gwbridge  proto kernel  scope link  src 172.18.0.1 linkdown 
192.168.19.0/24 dev enp0s3  proto kernel  scope link  src 192.168.19.121 
192.168.56.0/24 dev enp0s8  proto kernel  scope link  src 192.168.56.101 

配置docker连接flannel

编辑docker配置文件

*两台机器都做此操作
首先查看/run/flannel/subnet.env

root@host2:~# cat /run/flannel/subnet.env 
FLANNEL_NETWORK=10.2.0.0/16
FLANNEL_SUBNET=10.2.21.1/24
FLANNEL_MTU=1450
FLANNEL_IPMASQ=false

之后,编辑docker的配置文件vim /etc/systemd/system/docker.service.d/10-machine.conf,给ExecStart命令的参数中添加--bip和--mtu,分别对应subnet.env的 FLANNEL_SUBNETFLANNEL_MTU

root@host2:~# cat /etc/systemd/system/docker.service.d/10-machine.conf
[Service]
ExecStart=
ExecStart=/usr/bin/dockerd -H tcp://0.0.0.0:2376 -H unix:///var/run/docker.sock --storage-driver aufs --tlsverify --tlscacert /etc/docker/ca.pem --tlscert /etc/docker/server.pem --tlskey /etc/docker/server-key.pem --label provider=generic --bip=10.2.21.1/24 --mtu=1450
Environment=

重启docker

systemctl daemon-reload
systemctl restart docker.service

效果

对比路由表

# docker重启前
root@host3:~# ip r
default via 192.168.19.2 dev enp0s3 onlink 
10.2.21.0/24 via 10.2.21.0 dev flannel.1 onlink 
172.17.0.0/16 dev docker0  proto kernel  scope link  src 172.17.0.1 linkdown 
172.18.0.0/16 dev docker_gwbridge  proto kernel  scope link  src 172.18.0.1 linkdown 
192.168.19.0/24 dev enp0s3  proto kernel  scope link  src 192.168.19.122 
192.168.56.0/24 dev enp0s8  proto kernel  scope link  src 192.168.56.102 
root@host3:~# systemctl daemon-reload
root@host3:~# systemctl restart docker.service

# docker重启后
root@host3:~# ip r
default via 192.168.19.2 dev enp0s3 onlink 
10.2.21.0/24 via 10.2.21.0 dev flannel.1 onlink 
10.2.99.0/24 dev docker0  proto kernel  scope link  src 10.2.99.1 linkdown 
172.18.0.0/16 dev docker_gwbridge  proto kernel  scope link  src 172.18.0.1 linkdown 
192.168.19.0/24 dev enp0s3  proto kernel  scope link  src 192.168.19.122 
192.168.56.0/24 dev enp0s8  proto kernel  scope link  src 192.168.56.102 

flannel 没有创建新的 docker 网络,而是直接使用默认的 bridge 网络。同一主机的容器通过 docker0 连接,跨主机流量通过 flannel.1 转发。

将容器连接到flannel网络

root@host2:~# docker run -itd --name bbox1 busybox
root@host2:~# docker exec bbox1 ip r
default via 10.2.21.1 dev eth0 
10.2.21.0/24 dev eth0 scope link  src 10.2.21.2 

root@host3:~# docker run -itd --name bbox2 busybox
root@host3:~# docker exec bbox2 ip r
default via 10.2.99.1 dev eth0 
10.2.99.0/24 dev eth0 scope link  src 10.2.99.2

可以发现网络可通

# 10.2.99.2 为bbox2的ip
root@host2:~# docker exec bbox1 ping -c 1 10.2.99.2
PING 10.2.99.2 (10.2.99.2): 56 data bytes
64 bytes from 10.2.99.2: seq=0 ttl=62 time=0.604 ms

使用traceroute跟踪包

root@host2:~# docker exec bbox1 traceroute 10.2.99.2
traceroute to 10.2.99.2 (10.2.99.2), 30 hops max, 46 byte packets
 1  10.2.21.1 (10.2.21.1)  0.010 ms  0.216 ms  0.020 ms
 2  10.2.99.0 (10.2.99.0)  0.938 ms  2.731 ms  1.663 ms
 3  10.2.99.2 (10.2.99.2)  2.016 ms  0.647 ms  0.596 ms
  1. 发给docker0
  2. 之后发给flannel.1
  3. flannel.1 将数据包封装成 VxLAN,通过 enp0s8 发送给 host3,
  4. host3 收到包解封装,发现数据包目的地址为 10.2.99.2,根据路由表(下图)将数据包发送给 flannel.1,并通过 docker0 到达 bbox2

### dns不可用
无法通过主机名ping通机器。


felix0913
27 声望1 粉丝