/container-images

Primary LanguageShell

规范

必须谨慎操作，防止意外覆盖生产环境镜像

make -C openvswitch

buildx node设定

有两种办法

  • 一种是qemu usermode emulation，信赖tcg翻译，速度较慢，有内核版本依赖
  • 一种是native方法，要求有物理真机

binfmt_misc with qemu usermode emulation

docker/binfmt依赖binfmt_misc的F标志支持，要求内核版本至少为v4.8

mount -t binfmt_misc binfmt_misc /proc/sys/fs/binfmt_misc/
docker run --rm --privileged docker/binfmt:a7996909642ee92942dcd6cff44b9b95f08dad64

验证，除了register, status两个文件，应还有qemu-xxx

ls /proc/sys/fs/binfmt_misc/

验证，输出应为`aarch64```

docker run --rm -t arm64v8/alpine uname -m

use remote native builder node

为当前用户（执行docker client的用户）设定好ssh pubkey auth

docker context create node-aarch64 --docker "host=ssh://user@host:port"
docker context ls

设置docker buildx builder

buildx特性依赖docker 19.03

buildx node添加的顺序很重要，关系到后来build时--platform linux/aarch64,linux/amd64匹配到哪个node

docker buildx ls
docker buildx create --name buildx          node-aarch64
docker buildx create --name buildx --append default
docker buildx use buildx

验证，应包含一项DRIVER为docker-container

docker buildx ls