amazon-ec2 - 为什么 Apache airflow 使用命令 : 'airflow initdb' ? 失败

Question

我正在尝试在 AWS EC2 实例上安装 Airflow 。网络上的各种来源似乎很好地记录了该过程，但是，在“pip install” Airflow 之后我遇到了问题；执行命令 'airflow initdb' 时出现以下错误:

[2019-09-25 13:22:02,329] {__init__.py:51} INFO - Using executor SequentialExecutor
Traceback (most recent call last):
  File "/home/cloud-user/.local/bin/airflow", line 22, in <module>
    from airflow.bin.cli import CLIFactory
  File "/home/cloud-user/.local/lib/python2.7/site-packages/airflow/bin/cli.py", line 68, in <module>
    from airflow.www_rbac.app import cached_app as cached_app_rbac
  File "/home/cloud-user/.local/lib/python2.7/site-packages/airflow/www_rbac/app.py", line 26, in <module>
    from flask_appbuilder import AppBuilder, SQLA
  File "/home/cloud-user/.local/lib/python2.7/site-packages/flask_appbuilder/__init__.py", line 5, in <module>
    from .base import AppBuilder
  File "/home/cloud-user/.local/lib/python2.7/site-packages/flask_appbuilder/base.py", line 5, in <module>
    from .api.manager import OpenApiManager
  File "/home/cloud-user/.local/lib/python2.7/site-packages/flask_appbuilder/api/__init__.py", line 11, in <module>
    from marshmallow_sqlalchemy.fields import Related, RelatedList
  File "/home/cloud-user/.local/lib/python2.7/site-packages/marshmallow_sqlalchemy/__init__.py", line 1, in <module>
    from .schema import TableSchemaOpts, ModelSchemaOpts, TableSchema, ModelSchema
  File "/home/cloud-user/.local/lib/python2.7/site-packages/marshmallow_sqlalchemy/schema.py", line 101
    class TableSchema(ma.Schema, metaclass=TableSchemaMeta):

                                      ^

我的安装步骤如下:

sudo apt update -y
sudo apt install python-pip -y
pip install apache-airflow
export AIRFLOW_HOME=~/.local/bin
export PATH=$PATH:$AIRFLOW_HOME
cd $AIRFLOW_HOME
mkdir dags/
airflow initdb
airflow scheduler &
airflow webserver &

我也尝试按照一些在线说明安装 Postgres:

sudo apt-get install postgresql postgresql-contrib -y

或者:pip install apache-airflow[postgres]

关于可能需要的任何配置文件更改没有太多说明。

虽然我停留在 initdb 这一步，但最终目标是将数据库分离到它自己的 RDS 实例中。

我还阅读了一些文档，其中指出可以在虚拟环境中运行 Airflow ，但这是必需的吗？

Answer 1

那是因为 marshmallow-sqlalchemy 版本升级了，0.19.0 破坏了与 py2 和 py3.5 的兼容性

运行以下

pip install marshmallow-sqlalchemy==0.17.1

它应该可以工作。

关于 Airflow Jira 的相关问题:https://issues.apache.org/jira/browse/AIRFLOW-5430