airflow—执行器CeleryExecutor（3）

原创

刘远

修改于 2017-12-29 10:53:38

4.1K0

修改于 2017-12-29 10:53:38

文章被收录于专栏：刘远的专栏

CeleryExecutor可用于正式环境，使用 Celery 作为Task执行的引擎, 扩展性很好。这里使用rabbitmq作为celery的消息存储。

安装

在机器A和机器B上安装airflow

pip2 install airflow[celery] 
pip2 install airflow[rabbitmq]

注意：最新版本的celery（4.0.2）可能与rabbitmq的管理端不兼容，如果在rabbitmq管理端或用命令行工具显示多列时，报错如下

{error,{exit,{ucs,{bad_utf8_character_code}},
 [{xmerl_ucs,from_utf8,1,[{file,"xmerl_ucs.erl"},{line,185}]},
 {mochijson2,json_encode_string,2,[]},
 {mochijson2,'-json_encode_proplist/2-fun-0-',3,[]},
 {lists,foldl,3,[{file,"lists.erl"},{line,1197}]},
 {mochijson2,json_encode_proplist,2,[]},
 {mochijson2,'-json_encode_array/2-fun-0-',3,[]},
 {lists,foldl,3,[{file,"lists.erl"},{line,1197}]},
 {mochijson2,json_encode_array,2,[]}]}}

可以安装最新的celery-3.x.x的版本

 pip2 -U install celery==3.1.24

配置

设置executor

 # The executor class that airflow should use. Choices include
 # SequentialExecutor, LocalExecutor, CeleryExecutor
 #executor = SequentialExecutor
 executor = CeleryExecutor

设置broker_url

# The Celery broker URL. Celery supports RabbitMQ, Redis and experimentally
# a sqlalchemy database. Refer to the Celery documentation for more
# information.
#broker_url = sqla+mysql://airflow:airflow@localhost:3306/airflow
broker_url = amqp://username:passowrd@201401@host:5672/project

设置celery_result_backend

# Another key Celery setting
celery_result_backend = db+mysql://user:passowrd@host:port/airflow

启动

启动Workder

airflow worker -D

启动scheduler

airflow scheduler -D

增加一个DAG

将airflow例子example_bash_operator中的 schedule_interval 改为@once

dag = DAG(
    dag_id='example_bash_operator', default_args=args,
    #schedule_interval='0 0 * * *',
    schedule_interval="@once",
    dagrun_timeout=timedelta(minutes=60))

另存为文件 example_bash_operator.py

分别上传到机器A和机器B的dags_folder目录

$ tree
.
`-- airflow
 |-- airflow-scheduler.err
 |-- airflow-scheduler.log
 |-- airflow-scheduler.out
 |-- airflow.cfg
 |-- dags
 | |-- example_bash_operator.py

启动DAG

airflow trigger_dag example_bash_operator

查看业务日志

查看DAG任务

$ airflow list_tasks example_bash_operator 
also_run_this
run_after_loop
run_this_last
runme_0
runme_1
runme_2

机器A执行日志如下

$ tree
.
|-- also_run_this
| `-- 2017-04-13T20:42:35
|-- run_this_last
| `-- 2017-04-13T20:42:35
`-- runme_1
    `-- 2017-04-13T20:42:35

机器B执行日志如下

$ tree
.
|-- run_after_loop
| `-- 2017-04-13T20:42:35
|-- runme_0
| `-- 2017-04-13T20:42:35
`-- runme_2
   `-- 2017-04-13T20:42:35

从上面的日志文件可以看出，这个DAG的6个任务被分发到两台机器执行，每台机器执行3个任务。

业务日志的集中存储

airflow的log日志默认存储在文件中，也可以远程存储，配置如下

# Airflow can store logs remotely in AWS S3 or Google Cloud Storage. Users
# must supply a remote location URL (starting with either 's3://...' or
# 'gs://...') and an Airflow connection id that provides access to the storage
# location.
remote_base_log_folder =
remote_log_conn_id =
# Use server-side encryption for logs stored in S3
encrypt_s3_logs = False
# DEPRECATED option for remote log storage, use remote_base_log_folder instead!
s3_log_folder =

也可以通过logstach将日志搜集到Elasticsearch中存储

原创声明：本文系作者授权腾讯云开发者社区发表，未经许可，不得转载。

如有侵权，请联系 cloudcommunity@tencent.com 删除。

存储

rabbitmq

原创声明：本文系作者授权腾讯云开发者社区发表，未经许可，不得转载。

如有侵权，请联系 cloudcommunity@tencent.com 删除。

存储

rabbitmq

登录后参与评论

0 条评论

热度