原文地址： https://devcenter.heroku.com/articles/scheduled-jobs-custom-clock-processes-java-quartz-rabbitmq
Table of Contents
- Prerequisites
- Scheduling jobs with Quartz
- Queuing jobs with RabbitMQ
- Processing jobs
- Running on Heroku
- Further learning
There are numerous ways to schedule background jobs in Java applications. This article will teach you how to setup a Java application that uses the Quartz library along with RabbitMQ to create a scalable and reliable method of scheduling background jobs on Heroku.
Many of the common methods for background processing in Java advocate running background jobs within the same application as the web tier. This approach has scalability and reliability constraints.
A better approach is to move background jobs into separate processes so that the web tier is distinctly separate from the background processing tier. This allows the web tier to be exclusively for handling web requests. The scheduling of jobs should also be it’s own tier that puts jobs onto a queue. The worker processing tier can then be scaled independently from the rest of the application.
If you have questions about Java on Heroku, consider discussing them in the Java on Heroku forums.
The source for this article's reference application is available on GitHub.
Prerequisites
- The Heroku Toolbelt installed.
- Maven 3.0.4 installed.
- A Heroku user account. Signup is free and instant.
To clone the sample project to your local computer run:
$ git clone https://github.com/heroku/devcenter-java-quartz-rabbitmq.git
Cloning into devcenter-java-quartz-rabbitmq...
$ cd devcenter-java-quartz-rabbitmq/
Scheduling jobs with Quartz
A custom clock process will be used to create jobs and add them to a queue. To setup a custom clock process use the Quartz library. In Maven the dependency is declared with:
See the reference app's pom.xml for the full Maven build definition.
<dependency>
<groupId>org.quartz-scheduler</groupId>
<artifactId>quartz</artifactId>
<version>2.1.5</version> </dependency>
Now a Java application can be used to schedule jobs. Here is an example:
package com.heroku.devcenter;
import org.quartz.*;
import org.quartz.impl.StdSchedulerFactory; import org.slf4j.Logger; import org.slf4j.LoggerFactory; import static org.quartz.JobBuilder.newJob; import static org.quartz.SimpleScheduleBuilder.repeatSecondlyForever; import static org.quartz.TriggerBuilder.newTrigger; public class SchedulerMain { final static Logger logger = LoggerFactory.getLogger(SchedulerMain.class); public static void main(String[] args) throws Exception { Scheduler scheduler = StdSchedulerFactory.getDefaultScheduler(); scheduler.start(); JobDetail jobDetail = newJob(HelloJob.class).build(); Trigger trigger = newTrigger() .startNow() .withSchedule(repeatSecondlyForever(2)) .build(); scheduler.scheduleJob(jobDetail, trigger); } public static class HelloJob implements Job { @Override public void execute(JobExecutionContext jobExecutionContext) throws JobExecutionException { logger.info("HelloJob executed"); } } }
This simple example creates a HelloJob every two seconds that simply logs a message. Quartz has a very extensive API for creating Triggerschedules.
To test this application locally you can run the Maven build and then run the SchedulerMain Java class:
$ mvn package
INFO] Scanning for projects...
[INFO]
[INFO] ------------------------------------------------------------------------
[INFO] Building devcenter-java-quartz-rabbitmq 1.0-SNAPSHOT [INFO] ------------------------------------------------------------------------ $ java -cp target/classes:target/dependency/* com.heroku.devcenter.SchedulerMain ... 66 [main] INFO org.quartz.impl.StdSchedulerFactory - Quartz scheduler 'DefaultQuartzScheduler' initialized from default resource file in Quartz package: 'quartz.properties' 66 [main] INFO org.quartz.impl.StdSchedulerFactory - Quartz scheduler version: 2.1.5 66 [main] INFO org.quartz.core.QuartzScheduler - Scheduler DefaultQuartzScheduler_$_NON_CLUSTERED started. 104 [DefaultQuartzScheduler_Worker-1] INFO com.heroku.devcenter.SchedulerMain - HelloJob executed 2084 [DefaultQuartzScheduler_Worker-2] INFO com.heroku.devcenter.SchedulerMain - HelloJob executed
Press Ctrl-C to exit the app.
If the HelloJob actually did work itself then we would have a runtime bottleneck because we could not scale the scheduler while avoiding duplicate jobs being scheduled. Quartz does have a JDBC module that can use a database to prevent jobs from being duplicated but a simpler approach is to only run one instance of the scheduler and have the scheduled jobs added to a message queue where they can be processes in parallel by job worker processes.
Queuing jobs with RabbitMQ
RabbitMQ can be used as a message queue so the scheduler process can be used just to add jobs to a queue and worker processes can be used to grab jobs from the queue and process them. To add the RabbitMQ client library as a dependency in Maven specify the following in dependencies block of the pom.xml file:
<dependency>
<groupId>com.rabbitmq</groupId>
<artifactId>amqp-client</artifactId>
<version>2.8.2</version> </dependency>
If you want to test this locally then install RabbitMQ and set an environment variable that will provide the application the connection information to your RabbitMQ server.
On Windows:
$ set CLOUDAMQP_URL="amqp://guest:guest@localhost:5672/%2f"
On Mac/Linux:
$ export CLOUDAMQP_URL="amqp://guest:guest@localhost:5672/%2f"
The CLOUDAMQP_URL environment variable will be used by the scheduler and worker processes to connect to the shared message queue. This example uses that environment variable because that is the way theCloudAMQP Heroku Add-on will provide it’s connection information to the application.
The SchedulerMain class needs to be updated to add a new message onto a queue every time the HelloJob is executed. Here are the newSchedulerMain and HelloJob classes from the SchedulerMain.java file in the sample project:
package com.heroku.devcenter;
import com.rabbitmq.client.Channel;
import com.rabbitmq.client.Connection; import com.rabbitmq.client.ConnectionFactory; import com.rabbitmq.client.MessageProperties; import org.quartz.*; import org.quartz.impl.StdSchedulerFactory; import org.slf4j.Logger; import org.slf4j.LoggerFactory; import java.util.HashMap; import java.util.Map; import static org.quartz.JobBuilder.newJob; import static org.quartz.SimpleScheduleBuilder.repeatSecondlyForever; import static org.quartz.TriggerBuilder.newTrigger; public class SchedulerMain { final static Logger logger = LoggerFactory.getLogger(SchedulerMain.class); final static ConnectionFactory factory = new ConnectionFactory(); public static void main(String[] args) throws Exception { factory.setUri(System.getenv("CLOUDAMQP_URL")); Scheduler scheduler = StdSchedulerFactory.getDefaultScheduler(); scheduler.start(); JobDetail jobDetail = newJob(HelloJob.class).build(); Trigger trigger = newTrigger() .startNow() .withSchedule(repeatSecondlyForever(5)) .build(); scheduler.scheduleJob(jobDetail, trigger); } public static class HelloJob implements Job { @Override public void execute(JobExecutionContext jobExecutionContext) throws JobExecutionException { try { Connection connection = factory.newConnection(); Channel channel = connection.createChannel(); String queueName = "work-queue-1"; Map<String, Object> params = new HashMap<String, Object>(); params.put("x-ha-policy", "all"); channel.queueDeclare(queueName, true, false, false, params); String msg = "Sent at:" + System.currentTimeMillis(); byte[] body = msg.getBytes("UTF-8"); channel.basicPublish("", queueName, MessageProperties.PERSISTENT_TEXT_PLAIN, body); logger.info("Message Sent: " + msg); connection.close(); } catch (Exception e) { logger.error(e.getMessage(), e); } } } }
In this example every time the HelloJob is executed it adds a message onto a RabbitMQ message queue simply containing a String with the time the String was created. Running the updated SchedulerMainshould add a new message to the queue every 5 seconds.
Processing jobs
Next, create a Java application that will pull messages from the queue and handle them. This application will also use the RabbitFactoryUtilto get a connection to RabbitMQ from the CLOUDAMQP_URL environment variable. Here is the WorkerMain class from the WorkerMain.java file in the example project:
package com.heroku.devcenter;
import com.rabbitmq.client.Channel;
import com.rabbitmq.client.Connection; import com.rabbitmq.client.ConnectionFactory; import com.rabbitmq.client.QueueingConsumer; import org.slf4j.Logger; import org.slf4j.LoggerFactory; import java.util.Collections; import java.util.HashMap; import java.util.Map; public class WorkerMain { final static Logger logger = LoggerFactory.getLogger(WorkerMain.class); public static void main(String[] args) throws Exception { ConnectionFactory factory = new ConnectionFactory(); factory.setUri(System.getenv("CLOUDAMQP_URL")); Connection connection = factory.newConnection(); Channel channel = connection.createChannel(); String queueName = "work-queue-1"; Map<String, Object> params = new HashMap<String, Object>(); params.put("x-ha-policy", "all"); channel.queueDeclare(queueName, true, false, false, params); QueueingConsumer consumer = new QueueingConsumer(channel); channel.basicConsume(queueName, false, consumer); while (true) { QueueingConsumer.Delivery delivery = consumer.nextDelivery(); if (delivery != null) { String msg = new String(delivery.getBody(), "UTF-8"); logger.info("Message Received: " + msg); channel.basicAck(delivery.getEnvelope().getDeliveryTag(), false); } } } }
This class simply waits for new messages on the message queue and logs that it received them. You can run this example locally by doing a build and then running the WorkerMain class:
$ mvn package $ java -cp target/classes:target/dependency/* com.heroku.devcenter.WorkerMain
You can also run multiple instances of this example locally to see how the job processing can be horizontally distributed.
Running on Heroku
Now that you have everything working locally you can run this on Heroku. First declare the process model in a new file named Procfilecontaining:
scheduler: java $JAVA_OPTS -cp target/classes:target/dependency/* com.heroku.devcenter.SchedulerMain worker: java $JAVA_OPTS -cp target/classes:target/dependency/* com.heroku.devcenter.WorkerMain
This defines two process types that can be executed on Heroku; one named scheduler for the SchedulerMain app and one named workerfor the WorkerMain app.
To run on Heroku you will need to push a Git repository to Heroku containing the Maven build descriptor, source code, and Procfile. If you cloned the example project then you already have a Git repository. If you need to create a new git repository containing these files, run:
$ git init
$ git add src pom.xml Procfile
$ git commit -m init
Create a new application on Heroku from within the project’s root directory:
$ heroku create
Creating furious-cloud-2945... done, stack is cedar-14
http://furious-cloud-2945.herokuapp.com/ | git@heroku.com:furious-cloud-2945.git
Git remote heroku added
Then add the CloudAMQP add-on to your application:
$ heroku addons:add cloudamqp
Adding cloudamqp to furious-cloud-2945... done, v2 (free)
cloudamqp documentation available at: https://devcenter.heroku.com/articles/cloudamqp
Now push your Git repository to Heroku:
$ git push heroku master
Counting objects: 165, done.
Delta compression using up to 2 threads.
...
-----> Heroku receiving push -----> Java app detected ... -----> Discovering process types Procfile declares types -> scheduler, worker -----> Compiled slug size is 1.4MB -----> Launching... done, v5 http://furious-cloud-2945.herokuapp.com deployed to Heroku
This will run the Maven build for your project on Heroku and create a slug file containing the executable assets for your application. To run the application you will need to allocate dynos to run each process type. You should only allocate one dyno to run the scheduler process type to avoid duplicate job scheduling. You can allocate as many dynos as needed to the worker process type since it is event driven and parallelizable through the RabbitMQ message queue.
To allocate one dyno to the scheduler process type run:
$ heroku ps:scale scheduler=1
Scaling scheduler processes... done, now running 1
This should begin adding messages to the queue every five seconds. To allocate two dynos to the worker process type run:
$ heroku ps:scale worker=2
Scaling worker processes... done, now running 2
This will provision two dynos, each which will run the WorkerMain app and pull messages from the queue for processing. You can verify that this is happening by watching the Heroku logs for your application. To open a feed of your logs run:
$ heroku logs -t
2012-06-26T22:26:47+00:00 app[scheduler.1]: 100223 [DefaultQuartzScheduler_Worker-1] INFO com.heroku.devcenter.SchedulerMain - Message Sent: Sent at:1340749607126
2012-06-26T22:26:47+00:00 app[worker.2]: 104798 [main] INFO com.heroku.devcenter.WorkerMain - Message Received: Sent at:1340749607126
2012-06-26T22:26:52+00:00 app[scheduler.1]: 105252 [DefaultQuartzScheduler_Worker-2] INFO com.heroku.devcenter.SchedulerMain - Message Sent: Sent at:1340749612155
2012-06-26T22:26:52+00:00 app[worker.1]: 109738 [main] INFO com.heroku.devcenter.WorkerMain - Message Received: Sent at:1340749612155
In this example execution the scheduler creates 2 messages which are handled by the two different worker dynos (worker.1 and worker.2). This shows that the work is being scheduled and distributed correctly.
Further learning
This example application just shows the basics for architecting a scalable and reliable system for scheduling and processing background jobs. To learn more see:
