我要回家想一想~

這次遇到起 docker nginx 卻無法連線 

出現 curl: (56) Recv failure: Connection reset by peer

$ docker run -d -p 80:80 --name myweb nginx
$ curl http://localhost
curl: (56) Recv failure: Connection reset by peer

查了許久才發現是 docker veth 沒有正常 bridge 到 docker network interface

是因為 veth match 到 /lib/systemd/network/80-wired.network 這個設定

而卡在 configuring

可用 networkctl status -a 指令觀察狀態

$ networkctl status -a
● 8: veth23375aa
    Link File: n/a
    Network File: /lib/systemd/network/80-wired.network
    Type: ether
    State: degraded (configuring)
    Driver: veth
    HW Address: 1e:eb:7d:8b:a2:08
    MTU: 1500 (min: 68, max: 65535)
    QDisc: noqueue
    IPv6 Address Generation Mode: eui64
    Queue Length (Tx/Rx): 1/1
    Auto negotiation: no
    Speed: 10Gbps
    Duplex: full
    Port: tp
    Address: fe80::1ceb:7dff:fe8b:a208
    DHCP6 Client DUID: DUID-EN/Vendor:0000ab11ac337e4e1d50679b0000

/lib/systemd/network/80-wired.network 內容

ref: 

https://git.yoctoproject.org/poky/tree/meta/recipes-core/systemd/systemd-conf/wired.network?h=hardknott

比較 ubuntu 上 docker veth 正常會 match 到 /usr/lib/systemd/network/99-default.link

狀態會是 unmanaged

且 ubuntu 並沒有 80-wired.network 這個設定檔

$ networkctl status -a
● 8: vetheb18eef
    Link File: /usr/lib/systemd/network/99-default.link
    Network File: n/a
    Type: ether
    State: n/a (unmanaged)
    Driver: veth
    ...

解決方式：

建立 /etc/systemd/network/20-docker-veth.network 讓 veth 先 match 到並跳過管理

[Match]
Name=veth*
Driver=veth

[Link]
Unmanaged=true

然後重啟 systemd-networkd 

sudo systemctl restart systemd-networkd.service

ref: https://forums.docker.com/t/archlinux-container-veth-interfaces-being-assigned-to-wrong-bridge/107197/2