docker compose deploy fate cluster

news/2025/1/2 5:08:39/

官方文档写的不清晰

KubeFATE，用于生成部署脚本，链接

部署机就是下载了 KubeFATE的主机；运行机就是要安装fate容器的主机（部署机和运行机可以相同）

两个主机：并非必须 centos7，Ubuntu也行
Docker 版本 : 19.03.0+；Docker Compose 版本: 1.27.0+；
确保部署机可以ssh免密登录到两个运行节点主机上(如果部署机和运行机一样就不用管了)

$  vim docker-deploy/parties.conf
user=fate  # 改成 root 方便
dir=/data/projects/fate
party_list=(10000 9999) # 部署的 partyid，部署单边只填写一个
party_ip_list=(192.168.0.1 192.168.0.2) # 与partyid对应$  bash ./generate_config.sh
$  ls docker-deploy/outputs/
confs-10000.tar  confs-9999.tar  serving-10000.tar  serving-9999.tarbash ./docker_deploy.sh all --training  # 部署全部
bash ./docker_deploy.sh 9999 --training  # 部署单个

docker_deploy.sh 会将配置文件通过 scp 传到目标主机（所以需要免密登录），解压并执行容器创建运行

$  cd /data/projects/fate/confs-10000
$  docker compose ps
NAME                           IMAGE                                  COMMAND                  SERVICE             CREATED              STATUS                        PORTS
confs-10000-clustermanager-1   federatedai/eggroll:3.2.0-release     "/tini -- bash -c 'j…"   clustermanager      About a minute ago   Up About a minute             4670/tcp
confs-10000-fateflow-1         federatedai/fateflow:2.2.0-release    "/bin/bash -c 'set -…"   fateflow            About a minute ago   Up About a minute (healthy)   192.168.7.1:9360->9360/tcp, :::9360->9360/tcp, 192.168.7.1:9380->9380/tcp, :::9380->9380/tcp
confs-10000-mysql-1            mysql:8.0.28                          "docker-entrypoint.s…"   mysql               About a minute ago   Up About a minute             3306/tcp, 33060/tcp
confs-10000-nodemanager-1      federatedai/eggroll:3.2.0-release     "/tini -- bash -c 'j…"   nodemanager         About a minute ago   Up About a minute             4671/tcp
confs-10000-osx-1              federatedai/osx:2.2.0-release         "/tini -- bash -c 'j…"   osx                 About a minute ago   Up About a minute             192.168.7.1:9370->9370/tcp, :::9370->9370/tcp
confs-10000-fateboard-1        federatedai/fateboard:2.1.1-release   "sh -c 'java -Dsprin…"   fateboard           About a minute ago   Up About a minute             192.168.7.1:8080->8080/tcp$  docker compose exec fateflow bash