Java：安排作业按时间间隔运行

最近，我花了一些时间围绕Neo4j版本之间的滚动升级构建了一组测试，作为其中的一部分，我想记录升级发生时的群集状态。

测试的主线程会等待升级完成，因此我想每隔几秒钟登录另一个线程。 Alistair向我指出了ScheduledExecutorService ，该服务效果很好。

我结束了一个大致如下的测试：

public class MyUpgradeTest {@Testpublic void shouldUpgradeFromOneVersionToAnother() throws InterruptedException{ScheduledExecutorService scheduledExecutorService = Executors.newSingleThreadScheduledExecutor();scheduledExecutorService.scheduleAtFixedRate( new LogAllTheThings(), 0, 1, TimeUnit.SECONDS );Thread.sleep(10000);// do upgrade of clusterscheduledExecutorService.shutdown();}static class LogAllTheThings implements Runnable{@Overridepublic void run(){Date time = new Date( System.currentTimeMillis() );try{Map<String, Object> masterProperties = selectedProperties( client(), URI.create( "http://localhost:7474/" ) );System.out.println( String.format( "%s: %s", time, masterProperties ) );}catch ( Exception ignored ){ignored.printStackTrace();}}private static Client client(){DefaultClientConfig defaultClientConfig = new DefaultClientConfig();defaultClientConfig.getClasses().add( JacksonJsonProvider.class );return Client.create( defaultClientConfig );}public static Map<String, Object> selectedProperties( Client client, URI uri ){Map<String, Object> jmxProperties = new HashMap<String, Object>();ArrayNode transactionsProperties = jmxBean( client, uri, "org.neo4j/instance%3Dkernel%230%2Cname%3DTransactions" );addProperty( jmxProperties, transactionsProperties, "LastCommittedTxId" );ArrayNode kernelProperties = jmxBean( client, uri, "org.neo4j/instance%3Dkernel%230%2Cname%3DKernel" );addProperty( jmxProperties, kernelProperties, "KernelVersion" );ArrayNode haProperties = jmxBean( client, uri, "org.neo4j/instance%3Dkernel%230%2Cname%3DHigh+Availability" );addProperty( jmxProperties, haProperties, "Role" );addProperty( jmxProperties, haProperties, "InstanceId" );return jmxProperties;}private static void addProperty( Map<String, Object> jmxProperties, ArrayNode properties, String propertyName ){jmxProperties.put( propertyName, getProperty( properties, propertyName ) );}private static String getProperty( ArrayNode properties, String propertyName ){for ( JsonNode property : properties ){if ( property.get( "name" ).asText().equals( propertyName ) ){return property.get( "value" ).asText();}}throw new RuntimeException( "Could not find requested property: " + propertyName );}private static ArrayNode jmxBean( Client client, URI uri, String beanExtension ){ClientResponse clientResponse = client.resource( uri + "db/manage/server/jmx/domain/" + beanExtension ).accept( MediaType.APPLICATION_JSON ).get( ClientResponse.class );JsonNode transactionsBean = clientResponse.getEntity( JsonNode.class );return (ArrayNode) transactionsBean.get( 0 ).get( "attributes" );}}
}

LogAllTheThings每秒被调用一次，它记录Neo4j服务器作为JMX属性公开的KernelVersion，InstanceId，LastCommittedTxId和Role。

如果我们对本地Neo4j集群运行它，我们将看到以下内容：

Sun Nov 17 22:31:55 GMT 2013: {KernelVersion=Neo4j - Graph Database Kernel 2.0.0-M06, InstanceId=1, LastCommittedTxId=18, Role=master}
Sun Nov 17 22:31:56 GMT 2013: {KernelVersion=Neo4j - Graph Database Kernel 2.0.0-M06, InstanceId=1, LastCommittedTxId=18, Role=master}
Sun Nov 17 22:31:57 GMT 2013: {KernelVersion=Neo4j - Graph Database Kernel 2.0.0-M06, InstanceId=1, LastCommittedTxId=18, Role=master}
Sun Nov 17 22:31:58 GMT 2013: {KernelVersion=Neo4j - Graph Database Kernel 2.0.0-M06, InstanceId=1, LastCommittedTxId=18, Role=master}
Sun Nov 17 22:31:59 GMT 2013: {KernelVersion=Neo4j - Graph Database Kernel 2.0.0-M06, InstanceId=1, LastCommittedTxId=18, Role=master}
...
removed for brevity

下一步是同时获取集群所有成员的属性，然后我们可以引入另一个ExecutorService ，该线程的线程池为3，以便它将同时评估（至少接近）每台计算机：

static class LogAllTheThings implements Runnable{private ExecutorService executorService = Executors.newFixedThreadPool( 3 );@Overridepublic void run(){List<URI> machines = new ArrayList<>(  );machines.add(URI.create( "http://localhost:7474/" ));machines.add(URI.create( "http://localhost:7484/" ));machines.add(URI.create( "http://localhost:7494/" ));Map<URI, Future<Map<String, Object>>> futureJmxProperties = new HashMap<>(  );for ( final URI machine : machines ){Future<Map<String, Object>> futureProperties = executorService.submit( new Callable<Map<String, Object>>(){@Overridepublic Map<String, Object> call() throws Exception{try{return selectedProperties( client(), machine );}catch ( Exception ignored ){ignored.printStackTrace();return new HashMap<>();}}} );futureJmxProperties.put( machine, futureProperties );}Date time = new Date( System.currentTimeMillis() );System.out.println( time );for ( Map.Entry<URI, Future<Map<String, Object>>> uriFutureEntry : futureJmxProperties.entrySet() ){try{System.out.println( "==> " + uriFutureEntry.getValue().get() );}catch ( Exception ignored ){}}}// other methods the same as above}

我们将每个作业提交给ExecutorService，并返回一个Future ，并将其存储在地图中，然后再检索其结果。如果运行该命令，将看到以下输出：

Sun Nov 17 22:49:58 GMT 2013
==> {KernelVersion=Neo4j - Graph Database Kernel 2.0.0-M06, InstanceId=1, LastCommittedTxId=18, Role=master}
==> {KernelVersion=Neo4j - Graph Database Kernel 2.0.0-M06, InstanceId=2, LastCommittedTxId=18, Role=slave}
==> {KernelVersion=Neo4j - Graph Database Kernel 2.0.0-M06, InstanceId=3, LastCommittedTxId=18, Role=slave}
Sun Nov 17 22:49:59 GMT 2013
==> {KernelVersion=Neo4j - Graph Database Kernel 2.0.0-M06, InstanceId=1, LastCommittedTxId=18, Role=master}
==> {KernelVersion=Neo4j - Graph Database Kernel 2.0.0-M06, InstanceId=2, LastCommittedTxId=18, Role=slave}
==> {KernelVersion=Neo4j - Graph Database Kernel 2.0.0-M06, InstanceId=3, LastCommittedTxId=18, Role=slave}
Sun Nov 17 22:50:00 GMT 2013
==> {KernelVersion=Neo4j - Graph Database Kernel 2.0.0-M06, InstanceId=1, LastCommittedTxId=18, Role=master}
==> {KernelVersion=Neo4j - Graph Database Kernel 2.0.0-M06, InstanceId=2, LastCommittedTxId=18, Role=slave}
==> {KernelVersion=Neo4j - Graph Database Kernel 2.0.0-M06, InstanceId=3, LastCommittedTxId=18, Role=slave}...
removed for brevity

总体而言，这种方法效果很好，尽管我总是愿意学习更好的方法！

参考： Java：安排我们的JCG合作伙伴 Mark Needham在Mark Needham Blog博客上按时间间隔运行作业。