Bash 脚本自动化 Docker 镜像构建、推送和 EKS 服务更新-CSDN博客

在使用 Amazon EKS (Elastic Kubernetes Service) 管理容器化应用程序时,我们经常需要更新应用程序的 Docker 镜像,并重新部署服务。这个过程包括构建新的 Docker 镜像、推送到 Amazon ECR (Elastic Container Registry)、更新 EKS 节点组的 AMI (Amazon Machine Image) 以及重新部署 Kubernetes 服务。手动执行这些步骤不仅繁琐,而且容易出错。

为了简化这个过程,我编写了一个 Bash 脚本,可以自动完成上述所有步骤。该脚本具有以下功能:

构建新的 Docker 镜像并推送到 Amazon ECR
获取 BottleRocket AMI 的快照 ID
更新 EKS 节点组的 AMI 快照 ID
更新 Kubernetes Deployment 的镜像版本
发送 DingDing 通知,通知重启成功

脚本代码

#!/bin/bash

# 设置环境变量
ENV=test
USER=${ENV}-ai
HOST=10.0.9.136
SERVER_NAME=tree
IMAGE_REPO="820600610568.dkr.ecr.us-east-1.amazonaws.com/ai"
DINGDING_TOKEN=5ecc827755c91854fca04d34b58e255e3ac39475ee92bfebbc1159b6369d42b9

# 构建 Docker 镜像并推送到 ECR
build_image(){
  local VERSION="$(date +%Y%m%d%H%M%S)"
  export IMAGE="${IMAGE_REPO}/${SERVER_NAME}:${VERSION}"
  sed -i 's/env="dev"/env=os.getenv("ENV_NAME", "dev")/' run.py #不同环境sqs不一样
  #sed -i "s|{env}-light|${ENV}-${SERVER_NAME}|g" run.py
  sudo /usr/local/bin/aws ecr get-login-password --region us-east-1 | sudo docker login --username AWS --password-stdin "${IMAGE_REPO}"
  docker build -t ${IMAGE} .
  docker push ${IMAGE}
  docker rmi -f ${IMAGE}
}

# 获取 BottleRocket 镜像快照 ID
get_snapshot_id(){
  local SNAPSHOT_LOG="/root/stable-diffusion-on-eks/utils/bottlerocket-images-cache/${ENV}-${SERVER_NAME}-ai-log.txt"
  cd /root/stable-diffusion-on-eks/utils/bottlerocket-images-cache || return
  ./snapshot.sh "${IMAGE}" | tee "${SNAPSHOT_LOG}"
  export SNAPSHOT_ID=$(grep -oEi 'snap-[a-z0-9]+' "${SNAPSHOT_LOG}" | tail -1)
  if [ -z "$SNAPSHOT_ID" ]; then
    echo "new_snapshot_id is empty"
    send_dingding "failure"
    exit 1
  fi 
}

# 更新 EKS 服务
update_service(){
  # 更新 EC2NodeClass 的镜像快照 ID
  ssh "${USER}@${HOST}" "kubectl patch EC2NodeClass ${ENV}-${SERVER_NAME}-ai-gpu --type='json' -p '[{\"op\":\"replace\", \"path\":\"/spec/blockDeviceMappings/1/ebs/snapshotID\", \"value\":\"${SNAPSHOT_ID}\"}]'"
  # 更新 Deployment 的镜像版本
  ssh "${USER}@${HOST}" "kubectl set image deployment/${ENV}-${SERVER_NAME}-ai ${ENV}-${SERVER_NAME}-ai=${IMAGE}"

}

# 发送 DingDing 通知
send_dingding(){
  if [ "$1" == "success" ]; then
    color='#00EC00'
  else
    color='#FF0000'
  fi

  curl "https://oapi.dingtalk.com/robot/send?access_token=${DINGDING_TOKEN}" \
      -H 'Content-Type: application/json' \
      -d ' {"msgtype": "markdown","markdown": {"title":"Restart '"${ENV}-${SERVER_NAME}-ai"' '"$1"'","text": "<font face='"'黑体'"' color='"'$color'"'>Restart '"${ENV}-${SERVER_NAME}-ai tag:${VERSION} SNAPSHOT_ID:${SNAPSHOT_ID}"' '"$1"'</font>\n"}}'
}


# 执行主要步骤
build_image
get_snapshot_id
update_service
send_dingding "success"