Java: Schedule a job to run on a time interval

Recently I’ve spent some time building a set of tests around rolling upgrades between Neo4j versions and as part of that I wanted to log the state of the cluster as the upgrade was happening.

The main thread of the test blocks waiting until the upgrade is done so I wanted to log on another thread every few seconds. Alistair pointed me at the ScheduledExecutorService which worked quite nicely.

I ended up with a test which looked roughly like this:
 
 

01	public class MyUpgradeTest {

02

@Test

03	public void shouldUpgradeFromOneVersionToAnother() throws InterruptedException

04

{

05	ScheduledExecutorService scheduledExecutorService = Executors.newSingleThreadScheduledExecutor();

06	scheduledExecutorService.scheduleAtFixedRate( new LogAllTheThings(), 0, 1, TimeUnit.SECONDS );

07

08	Thread.sleep(10000);

09	// do upgrade of cluster

10	scheduledExecutorService.shutdown();

11

}

12

13	static class LogAllTheThings implements Runnable

14

{

15

@Override

16	public void run()

17

{

18	Date time = new Date( System.currentTimeMillis() );

19

20

try

21

{

22	Map<String, Object> masterProperties = selectedProperties( client(), URI.create( "http://localhost:7474/" ) );

23	System.out.println( String.format( "%s: %s", time, masterProperties ) );

24

}

25	catch ( Exception ignored )

26

{

27	ignored.printStackTrace();

28

}

29

}

30

31	private static Client client()

32

{

33	DefaultClientConfig defaultClientConfig = new DefaultClientConfig();

34	defaultClientConfig.getClasses().add( JacksonJsonProvider.class );

35	return Client.create( defaultClientConfig );

36

}

37

38	public static Map<String, Object> selectedProperties( Client client, URI uri )

39

{

40	Map<String, Object> jmxProperties = new HashMap<String, Object>();

41

42	ArrayNode transactionsProperties = jmxBean( client, uri,"org.neo4j/instance%3Dkernel%230%2Cname%3DTransactions" );

43	addProperty( jmxProperties, transactionsProperties, "LastCommittedTxId" );

44

45	ArrayNode kernelProperties = jmxBean( client, uri,"org.neo4j/instance%3Dkernel%230%2Cname%3DKernel" );

46	addProperty( jmxProperties, kernelProperties, "KernelVersion" );

47

48	ArrayNode haProperties = jmxBean( client, uri,"org.neo4j/instance%3Dkernel%230%2Cname%3DHigh+Availability" );

49	addProperty( jmxProperties, haProperties, "Role" );

50	addProperty( jmxProperties, haProperties, "InstanceId" );

51

52	return jmxProperties;

53

}

54

55	private static void addProperty( Map<String, Object> jmxProperties, ArrayNode properties, String propertyName )

56

{

57	jmxProperties.put( propertyName, getProperty( properties, propertyName ) );

58

}

59

60	private static String getProperty( ArrayNode properties, String propertyName )

61

{

62	for ( JsonNode property : properties )

63

{

64	if ( property.get( "name" ).asText().equals( propertyName ) )

65

{

66	return property.get( "value" ).asText();

67

}

68

}

69

70	throw new RuntimeException( "Could not find requested property: " + propertyName );

71

}

72

73	private static ArrayNode jmxBean( Client client, URI uri, String beanExtension )

74

{

75	ClientResponse clientResponse = client

76	.resource( uri + "db/manage/server/jmx/domain/" + beanExtension )

77	.accept( MediaType.APPLICATION_JSON )

78	.get( ClientResponse.class );

79

80	JsonNode transactionsBean = clientResponse.getEntity( JsonNode.class );

81	return (ArrayNode) transactionsBean.get( 0 ).get( "attributes" );

82

}

83

}

84

}

LogAllTheThings gets called once every second and it logs the KernelVersion, InstanceId, LastCommittedTxId and Role which Neo4j server exposes as JMX properties.

If we run that against a local Neo4j cluster we’d see something like the following:

1	Sun Nov 17 22:31:55 GMT 2013: {KernelVersion=Neo4j - Graph Database Kernel 2.0.0-M06, InstanceId=1, LastCommittedTxId=18, Role=master}

2	Sun Nov 17 22:31:56 GMT 2013: {KernelVersion=Neo4j - Graph Database Kernel 2.0.0-M06, InstanceId=1, LastCommittedTxId=18, Role=master}

3	Sun Nov 17 22:31:57 GMT 2013: {KernelVersion=Neo4j - Graph Database Kernel 2.0.0-M06, InstanceId=1, LastCommittedTxId=18, Role=master}

4	Sun Nov 17 22:31:58 GMT 2013: {KernelVersion=Neo4j - Graph Database Kernel 2.0.0-M06, InstanceId=1, LastCommittedTxId=18, Role=master}

5	Sun Nov 17 22:31:59 GMT 2013: {KernelVersion=Neo4j - Graph Database Kernel 2.0.0-M06, InstanceId=1, LastCommittedTxId=18, Role=master}

6

...

7	removed for brevity

The next step was to get the properties of all the members of the cluster at the same time and to do that we can introduce another ExecutorService which has a thread pool of 3 so that it will evaluate each machine (at least close to) simultaneously:

01	static class LogAllTheThings implements Runnable

02

{

03	private ExecutorService executorService = Executors.newFixedThreadPool( 3 );

04

05

@Override

06	public void run()

07

{

08	List<URI> machines = new ArrayList<>(  );

09	machines.add(URI.create( "http://localhost:7474/" ));

10	machines.add(URI.create( "http://localhost:7484/" ));

11	machines.add(URI.create( "http://localhost:7494/" ));

12

13	Map<URI, Future<Map<String, Object>>> futureJmxProperties = new HashMap<>(  );

14	for ( final URI machine : machines )

15

{

16	Future<Map<String, Object>> futureProperties = executorService.submit( newCallable<Map<String, Object>>()

17

{

18

@Override

19	public Map<String, Object> call() throws Exception

20

{

21

try

22

{

23	return selectedProperties( client(), machine );

24

}

25	catch ( Exception ignored )

26

{

27	ignored.printStackTrace();

28	return new HashMap<>();

29

}

30

}

31

} );

32

33	futureJmxProperties.put( machine, futureProperties );

34

}

35

36	Date time = new Date( System.currentTimeMillis() );

37	System.out.println( time );

38	for ( Map.Entry<URI, Future<Map<String, Object>>> uriFutureEntry : futureJmxProperties.entrySet() )

39

{

40

try

41

{

42	System.out.println( "==> " + uriFutureEntry.getValue().get() );

43

}

44	catch ( Exception ignored )

45

{

46

47

}

48

}

49

}

50

51	// other methods the same as above

52

}

We submit each job to the ExecutorService and receive back a Future which we store in a map before retrieving its result later on. If we run that we’ll see the following output:

01	Sun Nov 17 22:49:58 GMT 2013

02	==> {KernelVersion=Neo4j - Graph Database Kernel 2.0.0-M06, InstanceId=1, LastCommittedTxId=18, Role=master}

03	==> {KernelVersion=Neo4j - Graph Database Kernel 2.0.0-M06, InstanceId=2, LastCommittedTxId=18, Role=slave}

04	==> {KernelVersion=Neo4j - Graph Database Kernel 2.0.0-M06, InstanceId=3, LastCommittedTxId=18, Role=slave}

05	Sun Nov 17 22:49:59 GMT 2013

06	==> {KernelVersion=Neo4j - Graph Database Kernel 2.0.0-M06, InstanceId=1, LastCommittedTxId=18, Role=master}

07	==> {KernelVersion=Neo4j - Graph Database Kernel 2.0.0-M06, InstanceId=2, LastCommittedTxId=18, Role=slave}

08	==> {KernelVersion=Neo4j - Graph Database Kernel 2.0.0-M06, InstanceId=3, LastCommittedTxId=18, Role=slave}

09	Sun Nov 17 22:50:00 GMT 2013

10	==> {KernelVersion=Neo4j - Graph Database Kernel 2.0.0-M06, InstanceId=1, LastCommittedTxId=18, Role=master}

11	==> {KernelVersion=Neo4j - Graph Database Kernel 2.0.0-M06, InstanceId=2, LastCommittedTxId=18, Role=slave}

12	==> {KernelVersion=Neo4j - Graph Database Kernel 2.0.0-M06, InstanceId=3, LastCommittedTxId=18, Role=slave}

13

14

...

15	removed for brevity

Overall the approach works quite well although I’m always open to learning of a better way if there is one!
 

Reference: Java: Schedule a job to run on a time interval from our JCG partner Mark Needham at the Mark Needham Blogblog.

출처: http://www.javacodegeeks.com/2013/11/java-schedule-a-job-to-run-on-a-time-interval.html?utm_content=buffer7724f&utm_medium=social&utm_source=facebook.com&utm_campaign=buffer