使用 docker buildx 構建多 cpu 架構鏡像 - 環境配置篇

時間 2020-12-21

標籤 linux git github docker json ubuntu c# api 架構 dom 欄目 Docker 简体版

原文原文鏈接

使用 docker buildx 實現多平臺編譯

docker registry api v2 支持多 CPU 架構鏡像.linux

同時 harbor v2 也實現了 docker registry api v2 的支持.git

實現以後，將再也不有 cpu 架構困擾，一個 docker pull image:tag 將自動適配可適配的平臺。真正作到海納百川，一騎當千。github

0x01 準備

docker 開啓實驗模式
buildx 插件
qemu 模擬器
Linux kernel >= 4.8

linux 系統內核

因爲 binfmt_misc 機制開啓須要依賴 Linux kernel >= 4.8 。所以，在對 linux 系統操做選型上有必定要求。docker

建議使用 發行版 出場內核已經知足需求的操做系統。而不是選擇本身升級系統內核。json

ubuntu:18.04 LTS 及以上
debian:10 及以上

參考文章:ubuntu

build multi architecture docker image with buildx
開啓實驗模式

當前 buildx 仍是一個實驗模式, 如須要支持, 須要進行以下配置c#

開啓實驗模式。修改 /etc/docker/daemon.json, 增長 experimental 字段，以下。

# vi /etc/docker/daemon.json
{
    "experimental": true
}

安裝 buildx

將 buildx 放到 ~/.docker/cli-plugins/ 目錄下

# https://github.com/docker/buildx/blob/master/README.md#docker-ce

BUILDX_VERSION=v0.4.1
ARCH=$(uname -m)
[ "${ARCH}" == "x86_64" ] && ARCH=amd64
[ "${ARCH}" == "aarch64" ] && ARCH=arm64

mkdir -p ~/.docker/cli-plugins
wget -c https://github.com/docker/buildx/releases/download/${BUILDX_VERSION}/buildx-${BUILDX_VERSION}.linux-${ARCH} -O docker-buildx\
    && chmod +x docker-buildx   \
    && mv docker-buildx ~/.docker/cli-plugins/

安裝 qemu

下載並安裝 qumu 到 /usr/bin/qemu-$(uname -m)-staticapi

# https://github.com/multiarch/qemu-user-static#multiarchqemu-user-static-images

QEMU_VERSION=v5.0.0-2

wget -c https://github.com/multiarch/qemu-user-static/releases/download/${QEMU_VERSION}/qemu-$(uname -m)-static -O qemu-$(uname -m)-static \
    && chmod +x qemu-$(uname -m)-static     \
    && mv qemu-$(uname -m)-static /usr/local/bin/qemu-$(uname -m)-static

0x02 example

建立 buildx builder

## 1. 建立指定名稱及指定平臺的builder
docker buildx create --use --platform=linux/amd64,linux/arm64 --name localbuilder

### 建立隨機名稱的 默認參數 builder
# docker buildx create --use

建立 Dockerfile 和 bakefile

目錄結構以下架構

# tree
# .
# ├── alpine-bake
# │   └── alpine.Dockerfile
# └── bake.hcl

alpine-bake/alpine.Dockerfile

FROM alpine:3.12

ARG TARGETARCH
RUN echo "TARGETARCH is $TARGETARCH"

docker-bake.hcl

group "default" {
    targets = ["alpine"]
}

target "alpine" {
    context = "./alpine-bake"
    dockerfile = "alpine.Dockerfile"
    tags = ["docker.io/tangx/alpine:buildx-bake-hcl"]
    platforms = ["linux/amd64", "linux/arm64", "linux/arm/v6", "linux/arm/v7", "linux/s390x"]
    ## push to registry
    output = ["type=registry"]
    ## pull base image always
    pull = true
}

target "debian" {
    context = "./debian-bake"
    ## default: Dockerfile
    # dockerfile = "Dockerfile"  
    tags = ["docker.io/tangx/debian:buildx-bake-hcl"]

    platforms = ["linux/amd64", "linux/arm64", "linux/arm/v6", "linux/arm/v7", "linux/s390x"]

    ## push to registry
    output = ["type=registry"]
    ## pull base image always
    pull = true
}

使用 bake 執行編譯

docker buildx bake -f bake.hcl # -f alias to bake

當命令中省略 -f 參數時, 默認 bake 文件爲:dom

docker-bake.json
docker-bake.hcl
docker-compose.yaml
執行結果

能夠在 tangx/alpine:buildx-bake-hcl 看到，生成的鏡像，支持了 5個 cpu 架構。

執行過程

# docker buildx bake -f bake.hcl

[+] Building 46.9s (19/19) FINISHED
 => [internal] booting buildkit              3.3s
 => [internal] load build definition from alpine.Dockerfile            0.1s
 => [internal] load .dockerignore            0.1s
 => [linux/arm/v6 internal] load metadata for docker.io/library/alpine:3.12        13.1s
 => [linux/arm64 internal] load metadata for docker.io/library/alpine:3.12         10.4s
 => [linux/amd64 internal] load metadata for docker.io/library/alpine:3.12         13.1s
 => [linux/s390x internal] load metadata for docker.io/library/alpine:3.12         13.1s
 => [linux/arm/v7 internal] load metadata for docker.io/library/alpine:3.12        13.1s
 => [linux/amd64 1/2] FROM docker.io/library/alpine:3.12@sha256:c0e9560cda118f9ec63ddefb4a173a2b2a0347082d7dff7dc14272e7841a5b5a        6.1s
 => [linux/arm/v7 1/2] FROM docker.io/library/alpine:3.12@sha256:c0e9560cda118f9ec63ddefb4a173a2b2a0347082d7dff7dc14272e7841a5b5a       5.9s
 => [linux/arm/v6 1/2] FROM docker.io/library/alpine:3.12@sha256:c0e9560cda118f9ec63ddefb4a173a2b2a0347082d7dff7dc14272e7841a5b5a       6.0s
 => [linux/arm64 1/2] FROM docker.io/library/alpine:3.12@sha256:c0e9560cda118f9ec63ddefb4a173a2b2a0347082d7dff7dc14272e7841a5b5a        6.1s
 => [linux/s390x 1/2] FROM docker.io/library/alpine:3.12@sha256:c0e9560cda118f9ec63ddefb4a173a2b2a0347082d7dff7dc14272e7841a5b5a        6.3s
 => [linux/arm/v7 2/2] RUN echo "TARGETARCH is arm"       0.9s
 => [linux/arm/v6 2/2] RUN echo "TARGETARCH is arm"       0.8s
 => [linux/amd64 2/2] RUN echo "TARGETARCH is amd64"      0.7s
 => [linux/arm64 2/2] RUN echo "TARGETARCH is arm64"      0.5s
 => [linux/s390x 2/2] RUN echo "TARGETARCH is s390x"      0.4s
 => exporting to image         22.9s

使用命令行執行編譯

# docker buildx build --platform=linux/amd64,linux/arm64 --file alpine-bake/alpine.Dockerfile --tag tangx/alpine:multi-arch --push alpine-bake/

參數解釋

docker buildx build : 主命令及子命令
--platform=linux/amd64,linux/arm64 : 執行 cpu 架構
--file : 指定 Dockerfile 的文件及路徑。省略則表示當前目錄下的 Dockerfile
--tag : 鏡像名字及TAG
--push : 完成並推送到倉庫
alpine-bake : docker context 的目錄地址。

執行結果

略

在 github action 上使用 buildx

https://github.com/marketplace/actions/customizable-docker-buildx

troubleshoot

multiple platforms feature is currently not supported for docker driver

當前模式不支持 buildx : https://github.com/docker/cli/blob/master/experimental/README.md

# 開始實驗模式 
# ~/docker/daemon.json
{
    "experimental": true
}

# docker version -f ''
## true

auto-push is currently not implemented for docker driver
缺乏 builder

# 建立一個 builder 
docker buildx create --use # a random name

docker buildx create --user --name specified_name # specified name

failed to solve: rpc error: code = Unknown desc = failed to load LLB: runtime execution on platform linux/arm64 not supported
參考 qemu-user-static #getting-started 執行如下命令。若是正常返回，則表示成功

$ uname -m
x86_64

$ docker run --rm -t arm64v8/ubuntu uname -m
standard_init_linux.go:211: exec user process caused "exec format error"

$ docker run --rm --privileged multiarch/qemu-user-static --reset -p yes

$ docker run --rm -t arm64v8/ubuntu uname -m
aarch64