作者简介

王海龙,Rancher中国社区技术经理,负责Rancher中国技术社区的保护和经营。领有6年的云计算畛域教训,经验了OpenStack到Kubernetes的技术改革,无论底层操作系统Linux,还是虚拟化KVM或是Docker容器技术都有丰盛的运维和实践经验。

前 言

Kubernetes 在 Changelog 中发表自 Kubernetes 1.20 之后将弃用 Docker 作为容器运行时之后,containerd成为下一个容器运行时的热门选项。尽管 containerd 很早就曾经是 Docker 的一部分,然而纯正应用 containerd 还是给大家带来了诸多困扰,本文将介绍如何应用 containerd 配置镜像仓库和加速器。

本文将以K3s为例对containerd进行配置,如果您的环境未应用 K3s 而是应用的 Kubernetes,你也能够参考本文来配置 containerd 的镜像仓库,因为 containerd 的配置是通用的。

对于 K3s 和 containerd

K3s 是一个轻量级 Kubernetes 发行版,二进制大小小于100MB,所需内存不到Kubernetes的一半。K3s 为了升高资源耗费,将默认的 runtime 批改为 containerd,同时也内置了 Kubernetes CLI 工具 crictl和ctr。

K3s 默认的 containerd 配置文件目录为/var/lib/rancher/k3s/agent/etc/containerd/config.toml,但间接操作 containerd 的配置文件去设置镜像仓库或加速器相比于操作 docker 要简单许多。K3s 为了简化配置 containerd 镜像仓库的复杂度,K3s 会在启动时查看/etc/rancher/k3s/中是否存在 registries.yaml 文件,如果存在该文件,就会依据 registries.yaml 的内容转换为 containerd 的配置并存储到/var/lib/rancher/k3s/agent/etc/containerd/config.toml,从而升高了配置 containerd 镜像仓库的复杂度。

应用 K3s 配置公有镜像仓库

K3s 镜像仓库配置文件由两大部分组成:mirrorsconfigs:

  • Mirrors 是一个用于定义专用镜像仓库的名称和 endpoint 的指令
  • Configs 局部定义了每个 mirror 的 TLS 和证书配置。对于每个 mirror,你能够定义auth和/或tls

containerd 应用了相似 K8S 中 svc 与 endpoint 的概念,svc 能够了解为拜访名称,这个名称会解析到对应的 endpoint 上。也能够了解 mirror 配置就是一个反向代理,它把客户端的申请代理到 endpoint 配置的后端镜像仓库。mirror 名称能够随便填写,然而必须合乎IP或域名的定义规定。并且能够配置多个 endpoint,默认解析到第一个 endpoint,如果第一个 endpoint 没有返回数据,则主动切换到第二个 endpoint,以此类推。

比方以下配置示例:

mirrors:  "172.31.6.200:5000":    endpoint:      - "http://172.31.6.200:5000"  "rancher.ksd.top:5000":    endpoint:      - "http://172.31.6.200:5000"  "docker.io":    endpoint:      - "https://fogjl973.mirror.aliyuncs.com"      - "https://registry-1.docker.io"

能够通过 crictl pull 172.31.6.200:5000/library/alpinecrictl pull rancher.ksd.top:5000/library/alpine获取到镜像,但镜像都是从同一个仓库获取到的。

root@rancher-server:/etc/rancher/k3s# systemctl restart k3s.serviceroot@rancher-server:/etc/rancher/k3s# crictl pull 172.31.6.200:5000/library/alpineImage is up to date for sha256:a24bb4013296f61e89ba57005a7b3e52274d8edd3ae2077d04395f806b63d83eroot@rancher-server:/etc/rancher/k3s# crictl pull rancher.ksd.top:5000/library/alpineImage is up to date for sha256:a24bb4013296f61e89ba57005a7b3e52274d8edd3ae2077d04395f806b63d83eroot@rancher-server:/etc/rancher/k3s#

非平安(http)公有仓库配置

配置非平安(http)公有仓库,只须要在 endpoint 中指定 http 协定头的地址即可。

在没有 TLS 通信的状况下,须要为 endpoints 指定http:// ,否则将默认为 https。

无认证

如果你应用的是非平安(http)公有仓库,那么能够通过上面的参数来配置 K3s 连贯公有仓库:

root@ip-172-31-13-117:~# cat >> /etc/rancher/k3s/registries.yaml <<EOFmirrors:  "172.31.6.200:5000":    endpoint:      - "http://172.31.6.200:5000"EOFsystemctl restart k3s

而后能够通过 crictl 去 pull 镜像:

root@ip-172-31-13-117:~# crictl pull 172.31.6.200:5000/my-ubuntuImage is up to date for sha256:9499db7817713c4d10240ca9f5386b605ecff7975179f5a46e7ffd59fff462ee

接下来,在看一下 containerd 的配置,能够看到文件开端追加了如下配置:

root@ip-172-31-13-117:~# cat /var/lib/rancher/k3s/agent/etc/containerd/config.toml[plugins.cri.registry.mirrors][plugins.cri.registry.mirrors."172.31.6.200:5000"]  endpoint = ["http://172.31.6.200:5000"][plugins.cri.registry.mirrors."rancher.ksd.top:5000"]  endpoint = ["http://172.31.6.200:5000"]

有认证

如果你的非平安(http)公有仓库带有认证,那么能够通过上面的参数来配置 k3s 连贯公有仓库:

`root@ip-172-31-13-117:~# cat >> /etc/rancher/k3s/registries.yaml <<EOF
mirrors:
"35.182.134.80":

endpoint:  - "http://35.182.134.80"

configs:
"35.182.134.80":

auth:  username: admin # this is the registry username  password: Harbor12345 # this is the registry password

EOF
systemctl restart k3s

通过 crictl 去 pull 镜像:

root@ip-172-31-13-117:~# crictl pull 35.182.134.80/ksd/ubuntu:16.04
Image is up to date for sha256:9499db7817713c4d10240ca9f5386b605ecff7975179f5a46e7ffd59fff462ee

Containerd 配置文件开端追加了如下配置:```root@ip-172-31-13-117:~# cat /var/lib/rancher/k3s/agent/etc/containerd/config.toml[plugins.cri.registry.mirrors][plugins.cri.registry.mirrors."35.182.134.80"]  endpoint = ["http://35.182.134.80"][plugins.cri.registry.configs."35.182.134.80".auth]  username = "admin"  password = "Harbor12345"

平安(https)公有仓库配置

以下示例均启用了认证,所以每个示例都配置了configs.auth,如果理论环境未配置认证,删除configs.auth配置即可。

应用授信 ssl 证书

与非平安(http)公有仓库配置相似,只须要配置 endpoint 对应的仓库地址为 https 即可。

root@ip-172-31-13-117:~# cat >> /etc/rancher/k3s/registries.yaml <<EOFmirrors:  "harbor.kingsd.top":    endpoint:      - "https://harbor.kingsd.top"configs:  "harbor.kingsd.top":    auth:      username: admin # this is the registry username      password: Harbor12345 # this is the registry passwordEOFsystemctl restart k3s

通过 crictl 去 pull 镜像:

root@ip-172-31-13-117:~# crictl pull harbor.kingsd.top/ksd/ubuntu:16.04Image is up to date for sha256:9499db7817713c4d10240ca9f5386b605ecff7975179f5a46e7ffd59fff462ee

Containerd 配置文件开端追加了如下配置:

root@ip-172-31-13-117:~# cat /var/lib/rancher/k3s/agent/etc/containerd/config.toml[plugins.cri.registry.mirrors][plugins.cri.registry.mirrors."harbor.kingsd.top"]  endpoint = ["https://harbor.kingsd.top"][plugins.cri.registry.configs."harbor.kingsd.top".auth]  username = "admin"  password = "Harbor12345"

应用自签 ssl 证书

如果后端仓库应用的是自签名的 ssl 证书,那么须要配置 CA 证书 用于 ssl 证书的校验。

`root@ip-172-31-13-117:~# cat >> /etc/rancher/k3s/registries.yaml <<EOF
mirrors:
"harbor-ksd.kingsd.top":

endpoint:  - "https://harbor-ksd.kingsd.top"

configs:
"harbor-ksd.kingsd.top":

auth:  username: admin # this is the registry username  password: Harbor12345 # this is the registry passwordtls:  ca_file: /opt/certs/ca.crt

EOF
systemctl restart k3s

通过 crictl 去 pull 镜像:

root@ip-172-31-13-117:~# crictl pull harbor-ksd.kingsd.top/ksd/ubuntu:16.04
Image is up to date for sha256:9499db7817713c4d10240ca9f5386b605ecff7975179f5a46e7ffd59fff462ee

Containerd 配置文件开端追加了如下配置:

root@ip-172-31-13-117:~# cat /var/lib/rancher/k3s/agent/etc/containerd/config.toml
[plugins.cri.registry.mirrors]

[plugins.cri.registry.mirrors."harbor-ksd.kingsd.top"]
endpoint = ["https://harbor-ksd.kingsd.top"]

[plugins.cri.registry.configs."harbor-ksd.kingsd.top".auth]
username = "admin"
password = "Harbor12345"

[plugins.cri.registry.configs."harbor-ksd.kingsd.top".tls]
ca_file = "/opt/certs/ca.crt"

### ssl 双向认证如果镜像仓库配置了双向认证,那么须要为 containerd 配置 ssl 证书用于 镜像仓库对 containerd 做认证。

root@ip-172-31-13-117:~# cat >> /etc/rancher/k3s/registries.yaml <<EOF
mirrors:
"harbor-ksd.kingsd.top":

endpoint:  - "https://harbor-ksd.kingsd.top"

configs:
"harbor-ksd.kingsd.top":

auth:  username: admin # this is the registry username  password: Harbor12345 # this is the registry passwordtls:  ca_file: /opt/certs/ca.crt # path to the ca file used in the registry  cert_file: /opt/certs/harbor-ksd.kingsd.top.cert # path to the cert file used in the registry  key_file: /opt/certs/harbor-ksd.kingsd.top.key # path to the key file used in the registry

EOF
systemctl restart k3s

通过 crictl 去 pull 镜像:

root@ip-172-31-13-117:~# crictl pull harbor-ksd.kingsd.top/ksd/ubuntu:16.04
Image is up to date for sha256:9499db7817713c4d10240ca9f5386b605ecff7975179f5a46e7ffd59fff462ee

Containerd 配置文件开端追加了如下配置:```root@ip-172-31-13-117:~# cat /var/lib/rancher/k3s/agent/etc/containerd/config.toml[plugins.cri.registry.mirrors][plugins.cri.registry.mirrors."harbor-ksd.kingsd.top"]  endpoint = ["https://harbor-ksd.kingsd.top"][plugins.cri.registry.configs."harbor-ksd.kingsd.top".auth]  username = "admin"  password = "Harbor12345"[plugins.cri.registry.configs."harbor-ksd.kingsd.top".tls]  ca_file = "/opt/certs/ca.crt"  cert_file = "/opt/certs/harbor-ksd.kingsd.top.cert"  key_file = "/opt/certs/harbor-ksd.kingsd.top.key"

加速器配置

Containerd 与 docker 都有默认仓库,均为 docker.io 。如果配置中未指定 mirror 为 docker.io,containerd 后会主动加载 docker.io 配置。与 docker 不同的是,containerd 能够批改 docker.io 对应的 endpoint(默认为 https://registry-1.docker.io ) ,而 docker 无奈批改。

Docker 中能够通过 registry-mirrors 设置镜像减速地址。如果 pull 的镜像不带仓库地址(我的项目名+镜像名:tag),则会从默认镜像仓库去拉取镜像。如果配置了镜像减速地址,会先拜访镜像减速仓库,如果没有返回数据,再拜访默认的镜像仓库。

Containerd 目前没有间接配置镜像减速的性能,但 containerd 中能够批改 docker.io 对应的 endpoint,所以能够通过批改 endpoint 来实现镜像减速下载。因为 endpoint 是轮询拜访,所以能够给 docker.io 配置多个仓库地址来实现 减速地址+默认仓库地址。如下配置示例:

mirrors:  "docker.io":    endpoint:      - "https://fogjl973.mirror.aliyuncs.com"      - "https://registry-1.docker.io"EOFsystemctl restart k3s

Containerd 配置文件开端追加了如下配置:

root@ip-172-31-13-117:~# cat /var/lib/rancher/k3s/agent/etc/containerd/config.toml[plugins.cri.registry.mirrors][plugins.cri.registry.mirrors."docker.io"]  endpoint = ["https://fogjl973.mirror.aliyuncs.com", "https://registry-1.docker.io"]

残缺配置示例

  "192.168.50.119":    endpoint:      - "http://192.168.50.119"  "docker.io":    endpoint:      - "https://fogjl973.mirror.aliyuncs.com"      - "https://registry-1.docker.io"configs:  "192.168.50.119":    auth:      username: '' # this is the registry username      password: '' # this is the registry password    tls:      cert_file: '' # path to the cert file used in the registry      key_file: '' # path to the key file used in the registry      ca_file: '' # path to the ca file used in the registry  "docker.io":    auth:      username: '' # this is the registry username      password: '' # this is the registry password    tls:      cert_file: '' # path to the cert file used in the registry      key_file: '' # path to the key file used in the registry      ca_file: '' # path to the ca file used in the registry

参考链接

K3s公有镜像仓库配置:

https://docs.rancher.cn/docs/...

Containerd配置镜像仓库:

https://github.com/containerd...