mlflow详细安装部署

1、安装docker

# 安装工具
sudo yum install -y yum-utils
# 添加yum仓库配置
sudo yum-config-manager --add-repo https://download.docker.com/linux/centos/docker-ce.repo
sudo yum-config-manager --add-repo http://mirrors.aliyun.com/docker-ce/linux/centos/docker-ce.rep
# 更新yum缓存
sudo yum makecache fast
# 安装docker
yum install -y docker-ce docker-ce-cli containerd.io
# 查看安装状态
docker info
# 执行以下命令新建配置国内源加速
cat <<EOF > /etc/docker/daemon.json
{
  "registry-mirrors": [
    "https://docker.mirrors.ustc.edu.cn",
    "http://hub-mirror.c.163.com"
  ],
  "max-concurrent-downloads": 10,
  "log-driver": "json-file",
  "log-level": "warn",
  "log-opts": {
    "max-size": "10m",
    "max-file": "3"
    },
  "data-root": "/var/lib/docker"
}
EOF
# 启动服务
systemctl start docker
# 设置开机自启
systemctl enable docker
# 查看状态
systemctl status docker

2、Docker安装minio

# 拉取镜像
docker pull minio/minio
# 运行容器，如果9000端口被占用请修改
docker run -d -p 9000:9000 --name minio \
-e "MINIO_ACCESS_KEY=minio" \
-e "MINIO_SECRET_KEY=minio123" \
-v /opt/minio/data:/data \
-v /opt/minio/config:/root/.minio \
minio/minio server /data \
--console-address ":9000" --address ":9090"

3、访问minio界面

• 地址：<安装节点ip>:9000
• 用户名：minio
• 密码：minio123
• 创建Bucket：点击Create Bucket 输入名称 mlflow 并创建

4、安装Anaconda3

# 拉取包
wget https://mirrors.tuna.tsinghua.edu.cn/anaconda/archive/Anaconda3-2021.11-Linux-x86_64.sh
# 安装命令，一路回车+yes
bash Anaconda3-2021.11-Linux-x86_64.sh
# 将conda添加至环境变量
vim /etc/profole
# 在文件底部添加，注意根据实际修改的anaconda安装路径
export PATH=/root/anaconda3/bin:$PATH
# 使环境变量生效
source /etc/profile
# 修改为清华源，否则创建环境会因网络情况缓慢或者失败
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free/
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/msys2/
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/pytorch/
conda config --set show_channel_urls yes

5、创建激活conda环境

# 创建conda环境并安装python3.8，时间比较长请耐心等待
conda create -n mlflow-1.11.0 python==3.8
# 如果出现以下提示请耐心等待系统自动尝试下一个镜像源：Solving environment: failed with repodata from current_repodata.json, will retry with next repodata source.
# 注意开启新终端，执行以下命令激活conda环境
conda activate mlflow-1.11.0

6、安装所需依赖包

# 依次执行安装mlfow tracking server python需要的依赖包
pip install mlflow==1.11.0
pip install mysqlclient==1.4.6
pip install boto3

7、启动mlflow tracking server

# 暴露出minio url以及需要的ID和KEY，因为mlflow tracking server在上传模型文件时需要
export AWS_ACCESS_KEY_ID=minio
export AWS_SECRET_ACCESS_KEY=minio123
export MLFLOW_S3_ENDPOINT_URL=http://localhost:9000
# 在MySQL中创建库mlflow
create database if not exists `mlflow`;
# 启动mlflow server，注意根据实际情况修改mysql信息
mlflow server \
   --backend-store-uri mysql://<mysql用户名>:'<mysql密码>'@localhost/mlflow \
   --host 0.0.0.0 -p 5002 \
   --default-artifact-root s3://mlflow

8、启动可能出现的问题

# 问题一：
TypeError: Descriptors cannot not be created directly.
If this call came from a _pb2.py file, your generated code is out of date and must be regenerated with protoc >= 3.19.0.
If you cannot immediately regenerate your protos, some other possible workarounds are:
 1. Downgrade the protobuf package to 3.20.x or lower.
 2. Set PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=python (but this will use pure-Python parsing and will be much slower).
# 原因是protobuf版本问题，解决方案是在当前canda环境先卸载再指定版本安装
pip uninstall protobuf
pip install protobuf==3.19.0
# 问题二：
ImportError: libmysqlclient.so.20: cannot open shared object file: No such file or directory
# 原因是在/usr/lib64/中没有libmysqlclient.so.20，解决方案是找到当前系统中libmysqlclient.so.20的路径，然后创建一个软连接到/usr/lib64/libmysqlclient.so.20
[root@node1 ~]# find / -name "libmysqlclient.so.20"
/usr/local/mysql/lib/libmysqlclient.so.20
[root@node1 ~]# ln -s /usr/local/mysql/lib/libmysqlclient.so.20 /usr/lib64/libmysqlclient.so.20
# 问题三：
sqlalchemy.exc.OperationalError: (MySQLdb._exceptions.OperationalError) (2002, "Can't connect to local MySQL server through socket '/tmp/mysql.sock' (2)")
# 原因是找不到tmp下的mysql.sock文件，解决方案是需要找到mysql.sock所在的目录，然后建立/tmp/mysql.sock软连接到该文件上
[root@node1 ~]# find / -name "mysql.sock"
/var/lib/mysql/mysql.sock
[root@node1 ~]# ln -s /var/lib/mysql/mysql.sock /tmp/mysql.sock