Docker的原生overlay网络的实现原理
Posted
tags:
篇首语：本文由小常识网(cha138.com)小编为大家整理，主要介绍了Docker的原生overlay网络的实现原理相关的知识，希望对你有一定的参考价值。
系统环境manager node: CentOS Linux release 7.4.1708 (Core)
workr node: CentOS Linux release 7.5.1804 (Core)
Docker版本信息
manager node: Docker version 18.09.4, build d14af54266
worker node: Docker version 19.03.1, build 74b1e89
Docker Swarm系统环境
manager node: 192.168.246.194
worker node: 192.168.246.195
创建 docker swarm集群前的网络
manager node:
# docker network ls
NETWOrk ID NAME DRIVER SCOPE
e01d59fe00e5 bridge bridge local
15426f623c37 host host local
dd5d570ac60e none null localworker node:
# docker network ls
NETWOrk ID NAME DRIVER SCOPE
70ed15a24acd bridge bridge local
e2da5d928935 host host local
a7dbda3b96e8 none null local创建 docker swarm 集群
初始化 docker swarm 集群
manager node执行： docker swarm init
worker node执行: docker swarm join --token SWMTKN-1-0p3g6ijmphmw5xrikh9e3asg5n3yzan0eomnsx1xuvkovvgfsp-enrmg2lj1dejg5igmnpoaywr1 192.168.246.194:2377
说明??：
如果遗忘了docker swarm join的命令，可以使用下面命令查找：
（1）对于 worker 节点：docker swarm join-token worker
（2）对于 manager 节点：docker swarm join-token manager
查看集群节点信息
manager node:
# docker node ls
ID HOSTNAME STATUS AVAILABILITY MANAGER STATUS ENGINE VERSION
hplz9lawfpjx6fpz0j1bevocp MyTest03 Ready Active 19.03.1
q5af6b67bmho8z0d7**m2yy5j * mysql-nginx Ready Active Leader 18.09.4查看集群网络信息
manager node:
# docker network ls
NETWOrk ID NAME DRIVER SCOPE
e01d59fe00e5 bridge bridge local
7c90d1bf0f62 docker_gwbridge bridge local
15426f623c37 host host local
8lyfiluksqu0 ingress overlay swarm
dd5d570ac60e none null localworker node:
# docker network ls
NETWOrk ID NAME DRIVER SCOPE
70ed15a24acd bridge bridge local
985367037d3b docker_gwbridge bridge local
e2da5d928935 host host local
8lyfiluksqu0 ingress overlay swarm
a7dbda3b96e8 none null local说明??：
在docker swarm集群创建的开始，docker 会给每台host创建除了docker0以外的两个网络，分是bridge类型(docker_gwbridge网桥)和overlay类型(ingress)的网络，以及一个过渡的命名空间ingress_sbox，我们可以使用如下命令在 manager节点自建overlay网络，结果如下：
docker network create -d overlay uber-svc
再次查看 manager 和 worker 两台主机 docker swarm 集群网络：
manager node:
# docker network ls
NETWOrk ID NAME DRIVER SCOPE
e01d59fe00e5 bridge bridge local
7c90d1bf0f62 docker_gwbridge bridge local
15426f623c37 host host local
8lyfiluksqu0 ingress overlay swarm
dd5d570ac60e none null local
kzxwwwtunpqe uber-svc overlay swarm ===> 这个 network 就是我们刚新建的 uber-svcworker node:
# docker network ls
NETWOrk ID NAME DRIVER SCOPE
70ed15a24acd bridge bridge local
985367037d3b docker_gwbridge bridge local
e2da5d928935 host host local
8lyfiluksqu0 ingress overlay swarm
a7dbda3b96e8 none null local说明??：
我们会发现在 worker node上并没有 uber-svc 网络。这是因为只有当运行中的容器连接到覆盖网络的时候，该网络才变为可用状态。这种延迟生效策略通过减少网络梳理，提升了网络的扩展性。
查看网络命名空间信息
manager node:
# ip netns
1-8lyfiluksq (id: 0)
ingress_sbox (id: 1)worker node:
# ip netns
1-8lyfiluksq (id: 0)
ingress_sbox (id: 1)说明??：
（1）由于容器和overlay的网络的网络命名空间文件不再操作系统默认的/var/run/netns下，只能手动通过软连接的方式查看。ln -s /var/run/docker/netns /var/run/netns。
（2）有时候网络的网络命名空间名称前面会带上1-、2-等序号，有时候不带。但不影响网络的通信和操作。
查看网络IPAM（IP Address Management）信息
（1）ingress网络的IPAM( IP Address Management)分配如下：
manager node 和 worker node 是相同的：
# docker network inspect ingress
[
"Name": "ingress",
"Id": "8lyfiluksqu09jfdjndhj68hl",
"Created": "2019-09-09T17:59:06.326723762+08:00",
"Scope": "swarm",
"Driver": "overlay",
"EnableIPv6": false,
"IPAM":
"Driver": "default",
"Options": null,
"Config": [
"Subnet": "10.255.0.0/16", ===> ingress子网
"Gateway": "10.255.0.1" ===> ingress网关
（2）uber-svc自建的overlay会使用docker自动分配的IPAM：
# docker network inspect uber-svc
[
"Name": "uber-svc",
"Id": "kzxwwwtunpqeucnrhmirg6rhm",
"Created": "2019-09-09T10:14:06.606521342Z",
"Scope": "swarm",
"Driver": "overlay",
"EnableIPv6": false,
"IPAM":
"Driver": "default",
"Options": null,
"Config": [
"Subnet": "10.0.0.0/24", ===> uber-svc子网
"Gateway": "10.0.0.1" ===> uber-svc网关
Docker swarm 中的LB分为两种情况
（1）Ingress Load Balancing
（2）Internal Load Balancing
说明??：我们本节重点聊聊 LB 的第二种情况，即Internal Load Balancing～
定义 shell 脚本
在开始下面的实践之前，我们先编辑以下两个脚本。对于脚本的使用，我会给出具体实例～
第一个脚本docker_netns.sh:
#!/bin/bash
NAMESPACE=$1
if [[ -z $NAMESPACE ]];then
ls -1 /var/run/docker/netns/
exit 0
fi
NAMESPACE_FILE=/var/run/docker/netns/$NAMESPACE
if [[ ! -f $NAMESPACE_FILE ]];then
NAMESPACE_FILE=$(docker inspect -f ".NetworkSettings.SandboxKey" $NAMESPACE 2>/dev/null)
fi
if [[ ! -f $NAMESPACE_FILE ]];then
echo "Cannot open network namespace ‘$NAMESPACE‘: No such file or directory"
exit 1
fi
shift
if [[ $# -lt 1 ]]; then
echo "No command specified"
exit 1
fi
nsenter --net=$NAMESPACE_FILE $@说明??：
（1）该脚本通过指定容器id、name或者namespace快速进入容器的network namespace并执行相应的shell命令。
（2）如果不指定任何参数，则列举所有Docker容器相关的network namespaces。
执行脚本结果如下：
# sh docker_netns.sh ==> 列出所有的网络命名空间
1-ycqv46f5tl
8402c558c13c
ingress_sbox# sh docker_netns.sh deploy_nginx_nginx_1 ip r ==> 进入查看名为deploy_nginx_nginx_1容器ip信息
default via 172.18.0.1 dev eth0
172.18.0.0/16 dev eth0 proto kernel scope link src 172.18.0.2 # sh docker_netns.sh 8402c558c13c ip r ==> 进入和查看网络命名空间为8402c558c13c容器ip信息
default via 172.18.0.1 dev eth0
172.18.0.0/16 dev eth0 proto kernel scope link src 172.18.0.2 第二个脚本find_links.sh:
#!/bin/bash
DOCKER_NETNS_SCRIPT=./docker_netns.sh
IFINDEX=$1
if [[ -z $IFINDEX ]];then
for namespace in $($DOCKER_NETNS_SCRIPT);do
printf "\e[1;31m%s:\e[0m" $namespace
$DOCKER_NETNS_SCRIPT $namespace ip -c -o link
printf " "
done
else
for namespace in $($DOCKER_NETNS_SCRIPT);do
if $DOCKER_NETNS_SCRIPT $namespace ip -c -o link | grep -Pq "^$IFINDEX: ";then
printf "\e[1;31m%s:\e[0m" $namespace
$DOCKER_NETNS_SCRIPT $namespace ip -c -o link | grep -P "^$IFINDEX: ";
printf " "
fi
done
fi该脚本根据ifindex查找虚拟网络设备所在的namespace，脚本不同情况下执行结果如下：
# sh find_links.sh ==> 不指定ifindex，则列出所有namespaces的link设备。
# sh find_links.sh
1-3gt8phomoc:1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN mode DEFAULT group default qlen 1\ link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
2: ip_vti0@NONE: <NOARP> mtu 1480 qdisc noop state DOWN mode DEFAULT group default qlen 1\ link/ipip 0.0.0.0 brd 0.0.0.0
3: br0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1450 qdisc noqueue state UP mode DEFAULT group default \ link/ether e6:c5:04:ad:7b:31 brd ff:ff:ff:ff:ff:ff
74: vxlan0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1450 qdisc noqueue master br0 state UNKNOWN mode DEFAULT group default \ link/ether e6:c5:04:ad:7b:31 brd ff:ff:ff:ff:ff:ff link-netnsid 0
76: veth0@if75: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1450 qdisc noqueue master br0 state UP mode DEFAULT group default \ link/ether e6:fa:db:53:40:fd brd ff:ff:ff:ff:ff:ff link-netnsid 1
ingress_sbox:1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN mode DEFAULT group default qlen 1\ link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
2: ip_vti0@NONE: <NOARP> mtu 1480 qdisc noop state DOWN mode DEFAULT group default qlen 1\ link/ipip 0.0.0.0 brd 0.0.0.0
75: eth0@if76: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1450 qdisc noqueue state UP mode DEFAULT group default \ link/ether 02:42:0a:ff:00:02 brd ff:ff:ff:ff:ff:ff link-netnsid 0
78: eth1@if79: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP mode DEFAULT group default \ link/ether 02:42:ac:14:00:02 brd ff:ff:ff:ff:ff:ff link-netnsid 1# sh find_links.sh 76 ==> 指定ifindex=76
1-3gt8phomoc:76: veth0@if75: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1450 qdisc noqueue master br0 state UP mode DEFAULT group default \ link/ether e6:fa:db:53:40:fd brd ff:ff:ff:ff:ff:ff link-netnsid 1实战 -- Internal Load Balancing
部署一个 service 使用我们自己创建的 uber-svc 网络
docker service create --name uber-svc --network uber-svc -p 80:80 --replicas 2 nigelpoulton/tu-demo:v1
部署的这两个容器分别处于 manager 和 worker 节点上：
# docker service ls
ID NAME MODE REPLICAS IMAGE PORTS
pfnme5ytk59w uber-svc replicated 2/2 nigelpoulton/tu-demo:v1 *:80->80/tcp# docker service ps uber-svc
ID NAME IMAGE NODE DESIRED STATE CURRENT STATE ERROR PORTS
kh8zs9a2umwf uber-svc.1 nigelpoulton/tu-demo:v1 mysql-nginx Running Running 57 seconds ago
31p0rgg1f59w uber-svc.2 nigelpoulton/tu-demo:v1 MyTest03 Running Running 49 seconds ago说明??：
-p当然你也可以使用--publish代替-p，在这里的用意是将容器内部的服务暴露到host上，这样我们就可以访问这个services。
一般情况下我们在swarm中部署service后容器中的网络只有一张网卡使用的是docker0网络，当我们将服务发布出去后，swarm会做如下操作：
（1）给容器添加三块网卡eth0和eth1，eth2，eth0连接overlay类型网络名为ingress用于在不同主机间通信，eth1连接bridge类网络名为docker_gwbridge，用于让容器能访问外网。eth2连接到我们自己创建的mynet网络上，同样的作用也是用于容器之间的访问(区别于eth2网络存在dns解析即服务发现功能)。
（2）swarm各节点会利用ingress overlay网络负载均衡将服务发布到集群之外。
查看 uber-svc.1 容器和 uber-svc 网络命名空间的网卡情况
（1）先查看 uber-svc.1 容器
# docker ps
CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES
a2a763734e42 nigelpoulton/tu-demo:v1 "python app.py" About a minute ago Up About a minute 80/tcp uber-svc.1.kh8zs9a2umwf9cix381zr9x38（2）查看 uber-svc.1 容器中网卡情况
# sh docker_netns.sh uber-svc.1.kh8zs9a2umwf9cix381zr9x38 ip addr
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN group default 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
2: ip_vti0@NONE: <NOARP> mtu 1480 qdisc noop state DOWN group default qlen 1
link/ipip 0.0.0.0 brd 0.0.0.0
54: eth0@if55: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1450 qdisc noqueue state UP group default
link/ether 02:42:0a:ff:00:05 brd ff:ff:ff:ff:ff:ff link-netnsid 0
inet 10.255.0.5/16 brd 10.255.255.255 scope global eth0
valid_lft forever preferred_lft forever
56: eth2@if57: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP group default
link/ether 02:42:ac:13:00:03 brd ff:ff:ff:ff:ff:ff link-netnsid 2
inet 172.19.0.3/16 brd 172.19.255.255 scope global eth2
valid_lft forever preferred_lft forever
58: eth1@if59: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1450 qdisc noqueue state UP group default
link/ether 02:42:0a:00:00:03 brd ff:ff:ff:ff:ff:ff link-netnsid 1
inet 10.0.0.3/24 brd 10.0.0.255 scope global eth1
valid_lft forever preferred_lft forever当然，你也可以直接使用下面命令查看：
docker exec uber-svc.1.kh8zs9a2umwf9cix381zr9x38 ip addr
（3）查看 uber-svc 网络命名空间的网卡
# ip netns ==> 查看 manager 网络命名空间
d2feb68e3183 (id: 3)
1-kzxwwwtunp (id: 2)
lb_kzxwwwtun
1-8lyfiluksq (id: 0)
ingress_sbox (id: 1)# docker network ls ==> 查看 manager 集群网络
NETWOrk ID NAME DRIVER SCOPE
e01d59fe00e5 bridge bridge local
7c90d1bf0f62 docker_gwbridge bridge local
15426f623c37 host host local
8lyfiluksqu0 ingress overlay swarm
dd5d570ac60e none null local
kzxwwwtunpqe uber-svc overlay swarmsh docker_netns.sh 1-kzxwwwtunp ip addr ==> 查看 uber-svc 网络命名空间的网卡
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN group default 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
2: ip_vti0@NONE: <NOARP> mtu 1480 qdisc noop state DOWN group default qlen 1
link/ipip 0.0.0.0 brd 0.0.0.0
3: br0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1450 qdisc noqueue state UP group default
link/ether 3e:cb:12:d3:a3:cb brd ff:ff:ff:ff:ff:ff
inet 10.0.0.1/24 brd 10.0.0.255 scope global br0
valid_lft forever preferred_lft forever
51: vxlan0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1450 qdisc noqueue master br0 state UNKNOWN group default
link/ether e2:8e:35:4c:a3:7b brd ff:ff:ff:ff:ff:ff link-netnsid 0
53: veth0@if52: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1450 qdisc noqueue master br0 state UP group default
link/ether 3e:cb:12:d3:a3:cb brd ff:ff:ff:ff:ff:ff link-netnsid 1
59: veth1@if58: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1450 qdisc noqueue master br0 state UP group default
link/ether 9e:b4:8c:72:4e:74 brd ff:ff:ff:ff:ff:ff link-netnsid 2当然，你也可以使用下面的命令：
ip netns exec 1-kzxwwwtunp ip addr
# ip netns exec 1-kzxwwwtunp brctl show ==> 查看 uber-svc 网络命名空间的接口情况
bridge name bridge id STP enabled interfaces
br0 8000.3ecb12d3a3cb no veth0
veth1
vxlan0说明??：
<1> docker exec uber-svc.1.kh8zs9a2umwf9cix381zr9x38 ip addr这条命令可以看到 manager 节点上容器的网络有四张网卡，分别是：lo、eth0、eth1 和 eth2。
其中，eth1 对应的 veth pair为 uber-svc 网络中的veth1，eth2 对应的 veth pair为 host 上的vethef74971。
<2> ip netns exec 1-kzxwwwtunp brctl show查看 uber-svc 网络空间下网桥挂载情况可以看出veth1挂到了br0网桥上.
（4）查看 uber-svc 网络的vxlan-id
ip netns exec 1-kzxwwwtunp ip -o -c -d link show vxlan0
***** vxlan id 4097 *****uber-svc 网络命名空间与 service 容器之间的网络连接图
获取 ingress 命名空间信息
主要步骤如下：
（1）获取 ingress 的network信息
# docker network ls
NETWOrk ID NAME DRIVER SCOPE
8lyfiluksqu0 ingress overlay swarm（2）获取取 ingress 的命名空间信息
# ip netns
1-8lyfiluksq (id: 0)（3）获取 ingress 的命名空间中ip信息
# sh docker_netns.sh 1-8lyfiluksq ip addr
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN group default 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
2: ip_vti0@NONE: <NOARP> mtu 1480 qdisc noop state DOWN group default qlen 1
link/ipip 0.0.0.0 brd 0.0.0.0
3: br0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1450 qdisc noqueue state UP group default
link/ether 6e:5c:bd:c0:95:ea brd ff:ff:ff:ff:ff:ff
inet 10.255.0.1/16 brd 10.255.255.255 scope global br0
valid_lft forever preferred_lft forever
45: vxlan0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1450 qdisc noqueue master br0 state UNKNOWN group default
link/ether e6:f3:7a:00:85:e1 brd ff:ff:ff:ff:ff:ff link-netnsid 0
47: veth0@if46: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1450 qdisc noqueue master br0 state UP group default
link/ether fa:98:37:aa:83:2a brd ff:ff:ff:ff:ff:ff link-netnsid 1
55: veth1@if54: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1450 qdisc noqueue master br0 state UP group default
link/ether 6e:5c:bd:c0:95:ea brd ff:ff:ff:ff:ff:ff link-netnsid 2（4）获取 ingress 的命名空间中vxlan0的ID信息
# sh docker_netns.sh 1-8lyfiluksq ip -d link show vxlan0
***** vxlan id 4096 *****（5）获取 ingress 的命名空间中对应 veth pair 信息
# sh find_links.sh 46
ingress_sbox:46: eth0@if47: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1450 qdisc noqueue state UP mode DEFAULT group default \ link/ether 02:42:0a:ff:00:02 brd ff:ff:ff:ff:ff:ff link-netnsid 0ingress 网络命名空间与 service 容器之间的网络连接图
获取 ingress_sbox 网络命名空间信息
主要步骤如下：
（1）获取 ingress_sbox 的ip信息
# sh docker_netns.sh ingress_sbox ip addr
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN group default 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
2: ip_vti0@NONE: <NOARP> mtu 1480 qdisc noop state DOWN group default qlen 1
link/ipip 0.0.0.0 brd 0.0.0.0
46: eth0@if47: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1450 qdisc noqueue state UP group default
link/ether 02:42:0a:ff:00:02 brd ff:ff:ff:ff:ff:ff link-netnsid 0
inet 10.255.0.2/16 brd 10.255.255.255 scope global eth0
valid_lft forever preferred_lft forever
inet 10.255.0.4/32 brd 10.255.0.4 scope global eth0
valid_lft forever preferred_lft forever
49: eth1@if50: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP group default
link/ether 02:42:ac:13:00:02 brd ff:ff:ff:ff:ff:ff link-netnsid 1
inet 172.19.0.2/16 brd 172.19.255.255 scope global eth1
valid_lft forever preferred_lft forever（2）获取 ingress_sbox 的veth pair 接口信息
# sh find_links.sh 47
1-8lyfiluksq:47: veth0@if46: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1450 qdisc noqueue master br0 state UP mode DEFAULT group default \ link/ether fa:98:37:aa:83:2a brd ff:ff:ff:ff:ff:ff link-netnsid 1（3）获取 manager 主机上veth pair 接口信息
# ip link show
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN mode DEFAULT group default qlen 1
link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
2: ens37: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UP mode DEFAULT group default qlen 1000
link/ether 00:0c:29:25:8b:ac brd ff:ff:ff:ff:ff:ff
3: docker0: <NO-CARRIER,BROADCAST,MULTICAST,UP> mtu 1500 qdisc noqueue state DOWN mode DEFAULT group default
link/ether 02:42:cf:31:ee:03 brd ff:ff:ff:ff:ff:ff
14: ip_vti0@NONE: <NOARP> mtu 1480 qdisc noop state DOWN mode DEFAULT group default qlen 1
link/ipip 0.0.0.0 brd 0.0.0.0
48: docker_gwbridge: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP mode DEFAULT group default
link/ether 02:42:9c:aa:15:e6 brd ff:ff:ff:ff:ff:ff
50: vetheaa661b@if49: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue master docker_gwbridge state UP mode DEFAULT group default
link/ether 8a:3e:01:ab:db:75 brd ff:ff:ff:ff:ff:ff link-netnsid 1
57: vethef74971@if56: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue master docker_gwbridge state UP mode DEFAULT group default
link/ether 82:5c:65:e1:9c:e8 brd ff:ff:ff:ff:ff:ff link-netnsid 3ingress 网络命名空间与 ingree_sbox 网络命名空间之间的网络连接图
说明：swarm worker 节点上的情况与 manager 基本思路一样～
Swarm 总体的网络连接图
说明??：
（1）可以看到这里ingress_sbox和创建容器的ns共用一个ingress网络空间。
（2）通过使用docker exec [container ID/name] ip r会更加直观的看到网络流动情况，如下：
# docker exec uber-svc.1.kh8zs9a2umwf9cix381zr9x38 ip r
default via 172.19.0.1 dev eth2
10.0.0.0/24 dev eth1 proto kernel scope link src 10.0.0.3
10.255.0.0/16 dev eth0 proto kernel scope link src 10.255.0.5
172.19.0.0/16 dev eth2 proto kernel scope link src 172.19.0.3由此可知容器默认网关为172.19.0.1，也就是说容器是通过eth2出去的~
最后
关于 Docker Swarm 底层网络问题还有很多的知识点需要去探究，本节对最近学习到的docker network 做了一个基础总结，如有错误或不足，请各位大佬指正，感谢！
另：参考文档如有侵权，请及时与我联系，立删～。
最后，感谢开源，拥抱开源！
参考文档
以上是关于Docker的原生overlay网络的实现原理的主要内容，如果未能解决你的问题，请参考以下文章
overlay 如何实现跨主机通信？- 每天5分钟玩转 Docker 容器技术（52）