Easy Scheduler

Easy Scheduler for Big Data

** Design features: ** A distributed and easy-to-expand visual DAG workflow scheduling system. Dedicated to solving the complex dependencies in the data processing process, making the scheduling system out of the box in the data processing process. Its main objectives are as follows:  - Associate the Tasks according to the dependencies of the tasks in a DAG graph, which can visualize the running state of task in real time.  - Support for many task types: Shell, MR, Spark, SQL (mysql, postgresql, hive, sparksql), Python, Sub_Process, Procedure, etc.  - Support process scheduling, dependency scheduling, manual scheduling, manual pause/stop/recovery, support for failed retry/alarm, recovery from specified nodes, Kill task, etc.  - Support process priority, task priority and task failover and task timeout alarm/failure  - Support process global parameters and node custom parameter settings  - Support online upload/download of resource files, management, etc. Support online file creation and editing  - Support task log online viewing and scrolling, online download log, etc.  - Implement cluster HA, decentralize Master cluster and Worker cluster through Zookeeper  - Support online viewing of Master/Worker cpu load, memory, cpu  - Support process running history tree/gantt chart display, support task status statistics, process status statistics  - Support for complement  - Support for multi-tenant  - Support internationalization  - There are more waiting partners to explore

Comparison with similar scheduler systems

	EasyScheduler	Azkaban	Airflow
Stability
Single point of failure	Decentralized multi-master and multi-worker	Yes Single Web and Scheduler Combination Node	Yes. Single Scheduler
Additional HA requirements	Not required (HA is supported by itself)	DB	Celery / Dask / Mesos + Load Balancer + DB
Overload processing	Task queue mechanism, the number of schedulable tasks on a single machine can be flexibly configured, when too many tasks will be cached in the task queue, will not cause machine jam.	Jammed the server when there are too many tasks	Jammed the server when there are too many tasks
Easy to use
DAG Monitoring Interface	Visualization process defines key information such as task status, task type, retry times, task running machine, visual variables and so on at a glance.	Only task status can be seen	Can't visually distinguish task types
Visual process definition	Yes All process definition operations are visualized, dragging tasks to draw DAGs, configuring data sources and resources. At the same time, for third-party systems, the api mode operation is provided.	No DAG and custom upload via custom DSL	No DAG is drawn through Python code, which is inconvenient to use, especially for business people who can't write code.
Quick deployment	One-click deployment	Complex clustering deployment	Complex clustering deployment
Features
Suspend and resume	Support pause, recover operation	No Can only kill the workflow first and then re-run	No Can only kill the workflow first and then re-run
Whether to support multiple tenants	Users on easyscheduler can achieve many-to-one or one-to-one mapping relationship through tenants and Hadoop users, which is very important for scheduling large data jobs. " Supports traditional shell tasks, while supporting large data platform task scheduling: MR, Spark, SQL (mysql, postgresql, hive, sparksql), Python, Procedure, Sub_Process	No	No
Task type	Supports traditional shell tasks, and also support big data platform task scheduling: MR, Spark, SQL (mysql, postgresql, hive, sparksql), Python, Procedure, Sub_Process	shell、gobblin、hadoopJava、java、hive、pig、spark、hdfsToTeradata、teradataToHdfs	BashOperator、DummyOperator、MySqlOperator、HiveOperator、EmailOperator、HTTPOperator、SqlOperator
Compatibility	Support the scheduling of big data jobs like spark, hive, Mr. At the same time, it is more compatible with big data business because it supports multiple tenants.	Because it does not support multi-tenant, it is not flexible enough to use business in big data platform.	Because it does not support multi-tenant, it is not flexible enough to use business in big data platform.
Scalability
Whether to support custom task types	Yes	Yes	Yes
Is Cluster Extension Supported?	Yes The scheduler uses distributed scheduling, and the overall scheduling capability will increase linearly with the scale of the cluster. Master and Worker support dynamic online and offline.	Yes, but complicated Executor horizontal extend	Yes, but complicated Executor horizontal extend

System partial screenshot

Document

• Backend deployment documentation

• Front-end deployment documentation

• User manual

• Upgrade document

• Online Demo

More documentation please refer to EasyScheduler online documentation

Recent R&D plan

Work plan of Easy Scheduler: R&D plan, where In Develop card is the features of 1.1.0 version , TODO card is to be done (including feature ideas)

How to contribute code

Welcome to participate in contributing code, please refer to the process of submitting the code: https://github.com/analysys/EasyScheduler/blob/master/CONTRIBUTING.md

Thanks

Easy Scheduler uses a lot of excellent open source projects, such as google guava, guice, grpc, netty, ali bonecp, quartz, and many open source projects of apache, etc. It is because of the shoulders of these open source projects that the birth of the Easy Scheduler is possible. We are very grateful for all the open source software used! We also hope that we will not only be the beneficiaries of open source, but also be open source contributors, so we decided to contribute to easy scheduling and promised long-term updates. I also hope that partners who have the same passion and conviction for open source will join in and contribute to open source!

Help

The fastest way to get response from our developers is to submit issues, or add our wechat : 510570367