singh-harpreet-hs / scheduledwf

Schedule Conductor workflow is a scheduler as a service that runs in the cloud with Netflix conductor embedded in it. It runs as an extension module of conductor.

Schedule Conductor Workflows

Schedule Conductor workflow is a scheduler as a service that runs in the cloud with Netflix conductor embedded in it. It runs as an extension module of conductor.

Motivation

• In digital space there are many use cases that are solved by running schedulers. Some of the common cases are:
  1. publish site map for an e-commerce website.
  2. refresh cache everyday at a fix time.
  3. send an email notification at a scheduled time, etc.
• If your architecture is micro services based then you have only two options:
  1. either to add scheduling capability (like Quartz scheduler, Spring scheduler, etc) in the service that needs to schedule a task.
  2. or to setup a separate micro service, which will perform scheduling for required use cases. This service interact with other service for data needs through REST APIs.
• In the end we have unnecessary mesh of schedulers. After certain point of time such kind of setup is completely out of control.
• Solution to the problem is Schedule Conductor workflow:
  • This can be deployed in place of conductor server.
  • Embedded conductor provides service orchestration capability.
  • Extension module provides scheduling capability.
  • Conductor workflow with cron expression can be scheduled through REST API.
  • Scheduled job spawned internally starts a workflow at a scheduled time.
• Schedule Conductor workflow is extendable to work with all persistence stores supported by conductor.

Quickly use able in PRODUCTION:

• Scheduling module can be enabled with property conductor.additional.modules= io/github.jas34.scheduledwf.config.ScheduledWfServerModule
• Deploy scheduledwf-server instead of conductor-server.

You are done!

Schedule Conductor is compatible with Java 8 and has embedded Conductor v2.30.4

Architecture

High Level Architecture

API

• Expose REST API interface for scheduling a workflow with metadata definition and cron expression (Scheduled Workflow Metadata Management)
• Expose REST API interface for managing running schedulers (Scheduler Management)
• Read more about in section Scheduling And Managing A Workflow.

SERVICE

• Consists of manager to spawn a scheduler process.
• One stop to manage complete life cycle of schedulers.
• Read details in section Component Details.

STORE

• Currently implemented for MYSQL.
• Can be extended to other persistence stores offered by conductor.
• Get more details on DAO from section Persistence Layer.

Getting started

• Download jar from maven central ---TBD---
• Alternatively:
  • you can fork a branch from ---Link to master TBD---
  • run ./mvnw clean install
• Executable jar can be found at scheduledwf/scheduledwf-server/target/scheduledwf-server-{version}.jar
• start server with command: java -jar scheduledwf-server-{version}T.jar [PATH TO PROPERTY FILE] [log4j.properties file path]

Scheduling And Managing A Workflow

• REST operations for scheduling can be accessed on the conductor swagger at http://{host}:{port}
• Example: Let us schedule sample workflow with name testwf. Assuming workflow definition already exists in conductor server. You can use Cron Maker to generate cron expression.

  • From swagger useScheduled Workflow Metadata Management to add/update scheduling metadata.

    • POST /scheduling/metadata/scheduleWf: Schedule new workflow
    • GET /scheduling/metadata/scheduleWf: Get scheduling metadata of scheduled workflows.
    • GET /scheduling/metadata/scheduleWf/{name}: Get scheduling metadata of scheduled workflows by workflow name.
    • PUT /scheduling/metadata/scheduleWf/{name}: Update the status of scheduled workflow metadata.

  • From swagger useScheduler Management:

    • to search about schedule manager running on different nodes of cluster.
    • to search about scheduled jobs based upon scheduling metadata.
    • to search about different runs of scheduled jobs at scheduled time. The detailed data returned by IndexScheduledWfDAO

Runtime Model

Component Details

Scheduler Manager

• This component is meant to manage lifecycle of scheduled workflow.
• It takes lifecycle state decision for a scheduler with the help of scheduled jobs registry.
• It schedules the schedulers through scheduling assistant.
• It index the scheduling information through IndexScheduledWfDAO interface.

  Scheduled Jobs Registry

  • This registry act a single source of truth to know whether a particular workflow is required to be scheduled, paused or deleted.
  • It can be customized by implementing ScheduledProcessRegistry interface.
  • The default implementation can be found here io/github.jas34.scheduledwf.execution.DefaultScheduledProcessRegistry.
  • It reads scheduled workflow details through ScheduledWfExecutionDAO.

Scheduling Assistant

• This is an abstract layer for job scheduling.
• It comes with default implementation of DefaultWorkflowSchedulingAssistant.
• It:
  1. creates jobs.
  2. pause jobs.
  3. delete scheduled jobs.
• It contains factory WorkflowSchedulerFactory that returns an instance of WorkflowScheduler interface.
• The assistant can be customized with the implementation of WorkflowSchedulingAssistant interface.

  WorkflowSchedulerFactory

  • This is core abstraction to define Scheduled process of your choice. (WorkflowSchedulerFactory<T extends ScheduledProcess>)
  • ScheduledProcess is one of the granular entity that is expected to have reference to scheduled process/thread. Currently implemented as CronBasedScheduledProcess.
  • The default implementation can be found here DefaultWorkflowSchedulerFactory.
  • Currently returns CronBasedWorkflowScheduler.

Cron Based Workflow Scheduler (Jobs Scheduling)

• The scheduling capability is completely customizable by implementing WorkflowScheduler interface and by returning applicable instance from WorkflowSchedulerFactory.
• The default behaviour is to enable CronBasedWorkflowScheduler.
• This is composed of wisp scheduler. Wisp provides in memory scheduling capabilities. As a result, CronBasedWorkflowScheduler also schedules in memory schedulers using wisp.
• Each scheduled job is provided an instance of Runnable task through ScheduledTaskProvider interface.

  ScheduledTaskProvider

  • The default implementation of task provider is DefaultScheduledTaskProvider.
  • It creates fully flexible and customizable task for job before scheduling through getTask().
  • It performs indexing of workflow start executions through callback to IndexExecutionDataCallback.
  • It uses LockingService to prevent concurrent execution of same job on more than one server at scheduled time.

Locking Service

• LockingService is composed of LockProvider available in conductor.
• It contains a ExecutionPermitter:
  1. to get execution permit for a fix period of time through boolean issue(ScheduledTaskDef taskDef)
  2. to return back permit after use through void giveBack(ScheduledTaskDef taskDef).

  Execution Permitter

  • It consists of PermitDAO to persist the Permit for a fix period in the persistence store used by LockProvider.
  • This is currently supported for conductor lock mode:
    1. local_only
    2. redis
  • For any other type of data store, one can implement PermitDAO.

Persistence Layer

• The persistence layer has been designed in a way to be completely aligned with persistence architecture of conductor.
• This will be automatically enabled with conductor property db=MYSQL.
• Currently it has following DAO interfaces:

  ScheduledWfMetadataDAO

  • This is used to persist scheduling metadata definitions of workflow through swagger operations under Scheduled Workflow Metadata Management
  • This has been currently implemented for MYSQL only. One can implement ScheduledWfMetadataDAO for any other type of persistence store.

  ScheduledWfExecutionDAO

  • This is used to persist scheduling reference against scheduled workflow.
  • Currently implemented to operate in memory and we do not see any need to move to any other persistence layer.

  IndexScheduledWfDAO

  • This is used to index each run of a scheduled job with many additional details like:
    • status of last execution
    • last execution time
    • next execution time
    • node on which execution has happened, etc.
  • This has been currently implemented for MYSQL only. One can implement IndexScheduledWfDAO for Elastic Search persistence store.

Get Support

• This project is maintained by @jas34 and @sudshan as an open source application. Use github issue tracking for filing issues, ideas or support requests.
• We have a wide road map to add many features to this service through various customizable hooks described above.
• In case customization is an immediate for you, feel free to open an issue in github. We will consider that as a priority request.

Contributions

• Whether it is a small documentation correction, bug fix or new features, contributions are highly appreciated.
• Discuss new features and explore ideas with us through github issues.

License

TBA