温馨提示×

温馨提示×

您好,登录后才能下订单哦!

密码登录×
登录注册×
其他方式登录
点击 登录注册 即表示同意《亿速云用户服务条款》

Docker的原生overlay网络的实现原理

发布时间:2020-07-09 23:27:30 来源:网络 阅读:1092 作者:品鉴初心 栏目:云计算

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                local
worker 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                local
worker 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-svc
worker 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: eth2@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eth2,eth3,eth0连接overlay类型网络名为ingress用于在不同主机间通信,eth2连接bridge类网络名为docker_gwbridge,用于让容器能访问外网。eth3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eth3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: eth3@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 eth3
       valid_lft forever preferred_lft forever
58: eth2@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 eth2
       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             swarm
sh 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: veth2@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
                                             veth2
                                             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、eth2 和 eth3。
其中,eth2 对应的 veth pair为 uber-svc 网络中的veth2,eth3 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veth2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

Docker的原生overlay网络的实现原理

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: veth2@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 0

ingress 网络命名空间与 service 容器之间的网络连接图

Docker的原生overlay网络的实现原理

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: eth2@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 eth2
       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 3

ingress 网络命名空间与 ingree_sbox 网络命名空间之间的网络连接图

Docker的原生overlay网络的实现原理

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

Docker的原生overlay网络的实现原理

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 eth3
10.0.0.0/24 dev eth2  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 eth3  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eth3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

(1)Docker swarm中的LB和服务发现详解

(2)万字长文:聊聊几种主流Docker网络的实现原理

(3)Docker跨主机网络——overlay

(4)Docker 跨主机网络 overlay(十六)

(5)Docker overlay覆盖网络及VXLAN详解

向AI问一下细节

免责声明:本站发布的内容(图片、视频和文字)以原创、转载和分享为主,文章观点不代表本网站立场,如果涉及侵权请联系站长邮箱:is@yisu.com进行举报,并提供相关证据,一经查实,将立刻删除涉嫌侵权内容。

AI