Java Code Examples for org.apache.flink.streaming.util.AbstractStreamOperatorTestHarness#open()
The following examples show how to use
org.apache.flink.streaming.util.AbstractStreamOperatorTestHarness#open() .
You can vote up the ones you like or vote down the ones you don't like,
and go to the original project or source file by following the links above each example. You may check out the related API usage on the sidebar.
Example 1
Source File: RMQSinkTest.java From flink with Apache License 2.0 | 5 votes |
@Test public void testOpen() throws Exception { MockSerializationSchema<String> serializationSchema = new MockSerializationSchema<>(); RMQSink<String> producer = new RMQSink<>(rmqConnectionConfig, serializationSchema, publishOptions); AbstractStreamOperatorTestHarness<Object> testHarness = new AbstractStreamOperatorTestHarness<>( new StreamSink<>(producer), 1, 1, 0 ); testHarness.open(); assertThat("Open method was not called", serializationSchema.isOpenCalled(), is(true)); }
Example 2
Source File: FlinkKafkaConsumerBaseTest.java From flink with Apache License 2.0 | 5 votes |
@Test public void testOpen() throws Exception { MockDeserializationSchema<Object> deserializationSchema = new MockDeserializationSchema<>(); AbstractStreamOperatorTestHarness<Object> testHarness = createTestHarness( new DummyFlinkKafkaConsumer<>(new KafkaDeserializationSchemaWrapper<>(deserializationSchema)), 1, 0 ); testHarness.open(); assertThat("Open method was not called", deserializationSchema.isOpenCalled(), is(true)); }
Example 3
Source File: FlinkKafkaConsumerBaseMigrationTest.java From flink with Apache License 2.0 | 5 votes |
/** * Test restoring from an legacy empty state, when no partitions could be found for topics. */ @Test public void testRestoreFromEmptyStateNoPartitions() throws Exception { final DummyFlinkKafkaConsumer<String> consumerFunction = new DummyFlinkKafkaConsumer<>( Collections.singletonList("dummy-topic"), Collections.<KafkaTopicPartition>emptyList(), FlinkKafkaConsumerBase.PARTITION_DISCOVERY_DISABLED); StreamSource<String, DummyFlinkKafkaConsumer<String>> consumerOperator = new StreamSource<>(consumerFunction); final AbstractStreamOperatorTestHarness<String> testHarness = new AbstractStreamOperatorTestHarness<>(consumerOperator, 1, 1, 0); testHarness.setTimeCharacteristic(TimeCharacteristic.ProcessingTime); testHarness.setup(); // restore state from binary snapshot file testHarness.initializeState( OperatorSnapshotUtil.getResourceFilename( "kafka-consumer-migration-test-flink" + testMigrateVersion + "-empty-state-snapshot")); testHarness.open(); // assert that no partitions were found and is empty assertTrue(consumerFunction.getSubscribedPartitionsToStartOffsets() != null); assertTrue(consumerFunction.getSubscribedPartitionsToStartOffsets().isEmpty()); // assert that no state was restored assertTrue(consumerFunction.getRestoredState().isEmpty()); consumerOperator.close(); consumerOperator.cancel(); }
Example 4
Source File: ContinuousFileProcessingTest.java From flink with Apache License 2.0 | 4 votes |
@Test public void testFunctionRestore() throws Exception { String testBasePath = hdfsURI + "/" + UUID.randomUUID() + "/"; org.apache.hadoop.fs.Path path = null; long fileModTime = Long.MIN_VALUE; for (int i = 0; i < 1; i++) { Tuple2<org.apache.hadoop.fs.Path, String> file = createFileAndFillWithData(testBasePath, "file", i, "This is test line."); path = file.f0; fileModTime = hdfs.getFileStatus(file.f0).getModificationTime(); } TextInputFormat format = new TextInputFormat(new Path(testBasePath)); final ContinuousFileMonitoringFunction<String> monitoringFunction = createTestContinuousFileMonitoringFunction(format, FileProcessingMode.PROCESS_CONTINUOUSLY); StreamSource<TimestampedFileInputSplit, ContinuousFileMonitoringFunction<String>> src = new StreamSource<>(monitoringFunction); final AbstractStreamOperatorTestHarness<TimestampedFileInputSplit> testHarness = new AbstractStreamOperatorTestHarness<>(src, 1, 1, 0); testHarness.open(); final Throwable[] error = new Throwable[1]; final OneShotLatch latch = new OneShotLatch(); final DummySourceContext sourceContext = new DummySourceContext() { @Override public void collect(TimestampedFileInputSplit element) { latch.trigger(); } }; // run the source asynchronously Thread runner = new Thread() { @Override public void run() { try { monitoringFunction.run(sourceContext); } catch (Throwable t) { t.printStackTrace(); error[0] = t; } } }; runner.start(); // first condition for the source to have updated its state: emit at least one element if (!latch.isTriggered()) { latch.await(); } // second condition for the source to have updated its state: it's not on the lock anymore, // this means it has processed all the splits and updated its state. synchronized (sourceContext.getCheckpointLock()) {} OperatorSubtaskState snapshot = testHarness.snapshot(0, 0); monitoringFunction.cancel(); runner.join(); testHarness.close(); final ContinuousFileMonitoringFunction<String> monitoringFunctionCopy = createTestContinuousFileMonitoringFunction(format, FileProcessingMode.PROCESS_CONTINUOUSLY); StreamSource<TimestampedFileInputSplit, ContinuousFileMonitoringFunction<String>> srcCopy = new StreamSource<>(monitoringFunctionCopy); AbstractStreamOperatorTestHarness<TimestampedFileInputSplit> testHarnessCopy = new AbstractStreamOperatorTestHarness<>(srcCopy, 1, 1, 0); testHarnessCopy.initializeState(snapshot); testHarnessCopy.open(); Assert.assertNull(error[0]); Assert.assertEquals(fileModTime, monitoringFunctionCopy.getGlobalModificationTime()); hdfs.delete(path, false); }
Example 5
Source File: FromElementsFunctionTest.java From flink with Apache License 2.0 | 4 votes |
@Test public void testCheckpointAndRestore() { try { final int numElements = 10000; List<Integer> data = new ArrayList<Integer>(numElements); List<Integer> result = new ArrayList<Integer>(numElements); for (int i = 0; i < numElements; i++) { data.add(i); } final FromElementsFunction<Integer> source = new FromElementsFunction<>(IntSerializer.INSTANCE, data); StreamSource<Integer, FromElementsFunction<Integer>> src = new StreamSource<>(source); AbstractStreamOperatorTestHarness<Integer> testHarness = new AbstractStreamOperatorTestHarness<>(src, 1, 1, 0); testHarness.open(); final SourceFunction.SourceContext<Integer> ctx = new ListSourceContext<Integer>(result, 2L); final Throwable[] error = new Throwable[1]; // run the source asynchronously Thread runner = new Thread() { @Override public void run() { try { source.run(ctx); } catch (Throwable t) { error[0] = t; } } }; runner.start(); // wait for a bit Thread.sleep(1000); // make a checkpoint List<Integer> checkpointData = new ArrayList<>(numElements); OperatorSubtaskState handles = null; synchronized (ctx.getCheckpointLock()) { handles = testHarness.snapshot(566, System.currentTimeMillis()); checkpointData.addAll(result); } // cancel the source source.cancel(); runner.join(); // check for errors if (error[0] != null) { System.err.println("Error in asynchronous source runner"); error[0].printStackTrace(); fail("Error in asynchronous source runner"); } final FromElementsFunction<Integer> sourceCopy = new FromElementsFunction<>(IntSerializer.INSTANCE, data); StreamSource<Integer, FromElementsFunction<Integer>> srcCopy = new StreamSource<>(sourceCopy); AbstractStreamOperatorTestHarness<Integer> testHarnessCopy = new AbstractStreamOperatorTestHarness<>(srcCopy, 1, 1, 0); testHarnessCopy.setup(); testHarnessCopy.initializeState(handles); testHarnessCopy.open(); // recovery run SourceFunction.SourceContext<Integer> newCtx = new ListSourceContext<>(checkpointData); sourceCopy.run(newCtx); assertEquals(data, checkpointData); } catch (Exception e) { e.printStackTrace(); fail(e.getMessage()); } }
Example 6
Source File: FromElementsFunctionTest.java From flink with Apache License 2.0 | 4 votes |
@Test public void testCheckpointAndRestore() { try { final int numElements = 10000; List<Integer> data = new ArrayList<Integer>(numElements); List<Integer> result = new ArrayList<Integer>(numElements); for (int i = 0; i < numElements; i++) { data.add(i); } final FromElementsFunction<Integer> source = new FromElementsFunction<>(IntSerializer.INSTANCE, data); StreamSource<Integer, FromElementsFunction<Integer>> src = new StreamSource<>(source); AbstractStreamOperatorTestHarness<Integer> testHarness = new AbstractStreamOperatorTestHarness<>(src, 1, 1, 0); testHarness.open(); final SourceFunction.SourceContext<Integer> ctx = new ListSourceContext<Integer>(result, 2L); final Throwable[] error = new Throwable[1]; // run the source asynchronously Thread runner = new Thread() { @Override public void run() { try { source.run(ctx); } catch (Throwable t) { error[0] = t; } } }; runner.start(); // wait for a bit Thread.sleep(1000); // make a checkpoint List<Integer> checkpointData = new ArrayList<>(numElements); OperatorSubtaskState handles = null; synchronized (ctx.getCheckpointLock()) { handles = testHarness.snapshot(566, System.currentTimeMillis()); checkpointData.addAll(result); } // cancel the source source.cancel(); runner.join(); // check for errors if (error[0] != null) { System.err.println("Error in asynchronous source runner"); error[0].printStackTrace(); fail("Error in asynchronous source runner"); } final FromElementsFunction<Integer> sourceCopy = new FromElementsFunction<>(IntSerializer.INSTANCE, data); StreamSource<Integer, FromElementsFunction<Integer>> srcCopy = new StreamSource<>(sourceCopy); AbstractStreamOperatorTestHarness<Integer> testHarnessCopy = new AbstractStreamOperatorTestHarness<>(srcCopy, 1, 1, 0); testHarnessCopy.setup(); testHarnessCopy.initializeState(handles); testHarnessCopy.open(); // recovery run SourceFunction.SourceContext<Integer> newCtx = new ListSourceContext<>(checkpointData); sourceCopy.run(newCtx); assertEquals(data, checkpointData); } catch (Exception e) { e.printStackTrace(); fail(e.getMessage()); } }
Example 7
Source File: FlinkKinesisConsumerMigrationTest.java From flink with Apache License 2.0 | 4 votes |
@Test public void testRestoreWithEmptyState() throws Exception { final List<StreamShardHandle> initialDiscoveryShards = new ArrayList<>(TEST_STATE.size()); for (StreamShardMetadata shardMetadata : TEST_STATE.keySet()) { Shard shard = new Shard(); shard.setShardId(shardMetadata.getShardId()); SequenceNumberRange sequenceNumberRange = new SequenceNumberRange(); sequenceNumberRange.withStartingSequenceNumber("1"); shard.setSequenceNumberRange(sequenceNumberRange); initialDiscoveryShards.add(new StreamShardHandle(shardMetadata.getStreamName(), shard)); } final TestFetcher<String> fetcher = new TestFetcher<>( Collections.singletonList(TEST_STREAM_NAME), new TestSourceContext<>(), new TestRuntimeContext(true, 1, 0), TestUtils.getStandardProperties(), new KinesisDeserializationSchemaWrapper<>(new SimpleStringSchema()), null, initialDiscoveryShards); final DummyFlinkKinesisConsumer<String> consumerFunction = new DummyFlinkKinesisConsumer<>( fetcher, new KinesisDeserializationSchemaWrapper<>(new SimpleStringSchema())); StreamSource<String, DummyFlinkKinesisConsumer<String>> consumerOperator = new StreamSource<>(consumerFunction); final AbstractStreamOperatorTestHarness<String> testHarness = new AbstractStreamOperatorTestHarness<>(consumerOperator, 1, 1, 0); testHarness.setup(); testHarness.initializeState( OperatorSnapshotUtil.getResourceFilename( "kinesis-consumer-migration-test-flink" + testMigrateVersion + "-empty-snapshot")); testHarness.open(); consumerFunction.run(new TestSourceContext<>()); // assert that no state was restored assertTrue(consumerFunction.getRestoredState().isEmpty()); // although the restore state is empty, the fetcher should still have been registered the initial discovered shard; // furthermore, the discovered shard should be considered a newly created shard while the job wasn't running, // and therefore should be consumed from the earliest sequence number KinesisStreamShardState restoredShardState = fetcher.getSubscribedShardsState().get(0); assertEquals(TEST_STREAM_NAME, restoredShardState.getStreamShardHandle().getStreamName()); assertEquals(TEST_SHARD_ID, restoredShardState.getStreamShardHandle().getShard().getShardId()); assertFalse(restoredShardState.getStreamShardHandle().isClosed()); assertEquals(SentinelSequenceNumber.SENTINEL_EARLIEST_SEQUENCE_NUM.get(), restoredShardState.getLastProcessedSequenceNum()); consumerOperator.close(); consumerOperator.cancel(); }
Example 8
Source File: FlinkKinesisConsumerMigrationTest.java From flink with Apache License 2.0 | 4 votes |
@Test public void testRestoreWithReshardedStream() throws Exception { final List<StreamShardHandle> initialDiscoveryShards = new ArrayList<>(TEST_STATE.size()); for (StreamShardMetadata shardMetadata : TEST_STATE.keySet()) { // setup the closed shard Shard closedShard = new Shard(); closedShard.setShardId(shardMetadata.getShardId()); SequenceNumberRange closedSequenceNumberRange = new SequenceNumberRange(); closedSequenceNumberRange.withStartingSequenceNumber("1"); closedSequenceNumberRange.withEndingSequenceNumber("1087654321"); // this represents a closed shard closedShard.setSequenceNumberRange(closedSequenceNumberRange); initialDiscoveryShards.add(new StreamShardHandle(shardMetadata.getStreamName(), closedShard)); // setup the new shards Shard newSplitShard1 = new Shard(); newSplitShard1.setShardId(KinesisShardIdGenerator.generateFromShardOrder(1)); SequenceNumberRange newSequenceNumberRange1 = new SequenceNumberRange(); newSequenceNumberRange1.withStartingSequenceNumber("1087654322"); newSplitShard1.setSequenceNumberRange(newSequenceNumberRange1); newSplitShard1.setParentShardId(TEST_SHARD_ID); Shard newSplitShard2 = new Shard(); newSplitShard2.setShardId(KinesisShardIdGenerator.generateFromShardOrder(2)); SequenceNumberRange newSequenceNumberRange2 = new SequenceNumberRange(); newSequenceNumberRange2.withStartingSequenceNumber("2087654322"); newSplitShard2.setSequenceNumberRange(newSequenceNumberRange2); newSplitShard2.setParentShardId(TEST_SHARD_ID); initialDiscoveryShards.add(new StreamShardHandle(shardMetadata.getStreamName(), newSplitShard1)); initialDiscoveryShards.add(new StreamShardHandle(shardMetadata.getStreamName(), newSplitShard2)); } final TestFetcher<String> fetcher = new TestFetcher<>( Collections.singletonList(TEST_STREAM_NAME), new TestSourceContext<>(), new TestRuntimeContext(true, 1, 0), TestUtils.getStandardProperties(), new KinesisDeserializationSchemaWrapper<>(new SimpleStringSchema()), null, initialDiscoveryShards); final DummyFlinkKinesisConsumer<String> consumerFunction = new DummyFlinkKinesisConsumer<>( fetcher, new KinesisDeserializationSchemaWrapper<>(new SimpleStringSchema())); StreamSource<String, DummyFlinkKinesisConsumer<String>> consumerOperator = new StreamSource<>(consumerFunction); final AbstractStreamOperatorTestHarness<String> testHarness = new AbstractStreamOperatorTestHarness<>(consumerOperator, 1, 1, 0); testHarness.setup(); testHarness.initializeState( OperatorSnapshotUtil.getResourceFilename( "kinesis-consumer-migration-test-flink" + testMigrateVersion + "-snapshot")); testHarness.open(); consumerFunction.run(new TestSourceContext<>()); // assert that state is correctly restored assertNotEquals(null, consumerFunction.getRestoredState()); assertEquals(1, consumerFunction.getRestoredState().size()); assertEquals(TEST_STATE, removeEquivalenceWrappers(consumerFunction.getRestoredState())); // assert that the fetcher is registered with all shards, including new shards assertEquals(3, fetcher.getSubscribedShardsState().size()); KinesisStreamShardState restoredClosedShardState = fetcher.getSubscribedShardsState().get(0); assertEquals(TEST_STREAM_NAME, restoredClosedShardState.getStreamShardHandle().getStreamName()); assertEquals(TEST_SHARD_ID, restoredClosedShardState.getStreamShardHandle().getShard().getShardId()); assertTrue(restoredClosedShardState.getStreamShardHandle().isClosed()); assertEquals(TEST_SEQUENCE_NUMBER, restoredClosedShardState.getLastProcessedSequenceNum()); KinesisStreamShardState restoredNewSplitShard1 = fetcher.getSubscribedShardsState().get(1); assertEquals(TEST_STREAM_NAME, restoredNewSplitShard1.getStreamShardHandle().getStreamName()); assertEquals(KinesisShardIdGenerator.generateFromShardOrder(1), restoredNewSplitShard1.getStreamShardHandle().getShard().getShardId()); assertFalse(restoredNewSplitShard1.getStreamShardHandle().isClosed()); // new shards should be consumed from the beginning assertEquals(SentinelSequenceNumber.SENTINEL_EARLIEST_SEQUENCE_NUM.get(), restoredNewSplitShard1.getLastProcessedSequenceNum()); KinesisStreamShardState restoredNewSplitShard2 = fetcher.getSubscribedShardsState().get(2); assertEquals(TEST_STREAM_NAME, restoredNewSplitShard2.getStreamShardHandle().getStreamName()); assertEquals(KinesisShardIdGenerator.generateFromShardOrder(2), restoredNewSplitShard2.getStreamShardHandle().getShard().getShardId()); assertFalse(restoredNewSplitShard2.getStreamShardHandle().isClosed()); // new shards should be consumed from the beginning assertEquals(SentinelSequenceNumber.SENTINEL_EARLIEST_SEQUENCE_NUM.get(), restoredNewSplitShard2.getLastProcessedSequenceNum()); consumerOperator.close(); consumerOperator.cancel(); }
Example 9
Source File: FlinkKafkaConsumerBaseMigrationTest.java From flink with Apache License 2.0 | 4 votes |
/** * Test restoring from an empty state taken using a previous Flink version, when some partitions could be * found for topics. */ @Test public void testRestoreFromEmptyStateWithPartitions() throws Exception { final List<KafkaTopicPartition> partitions = new ArrayList<>(PARTITION_STATE.keySet()); final DummyFlinkKafkaConsumer<String> consumerFunction = new DummyFlinkKafkaConsumer<>(TOPICS, partitions, FlinkKafkaConsumerBase.PARTITION_DISCOVERY_DISABLED); StreamSource<String, DummyFlinkKafkaConsumer<String>> consumerOperator = new StreamSource<>(consumerFunction); final AbstractStreamOperatorTestHarness<String> testHarness = new AbstractStreamOperatorTestHarness<>(consumerOperator, 1, 1, 0); testHarness.setTimeCharacteristic(TimeCharacteristic.ProcessingTime); testHarness.setup(); // restore state from binary snapshot file testHarness.initializeState( OperatorSnapshotUtil.getResourceFilename( "kafka-consumer-migration-test-flink" + testMigrateVersion + "-empty-state-snapshot")); testHarness.open(); // the expected state in "kafka-consumer-migration-test-flink1.2-snapshot-empty-state"; // all new partitions after the snapshot are considered as partitions that were created while the // consumer wasn't running, and should start from the earliest offset. final HashMap<KafkaTopicPartition, Long> expectedSubscribedPartitionsWithStartOffsets = new HashMap<>(); for (KafkaTopicPartition partition : PARTITION_STATE.keySet()) { expectedSubscribedPartitionsWithStartOffsets.put(partition, KafkaTopicPartitionStateSentinel.EARLIEST_OFFSET); } // assert that there are partitions and is identical to expected list assertTrue(consumerFunction.getSubscribedPartitionsToStartOffsets() != null); assertTrue(!consumerFunction.getSubscribedPartitionsToStartOffsets().isEmpty()); assertEquals(expectedSubscribedPartitionsWithStartOffsets, consumerFunction.getSubscribedPartitionsToStartOffsets()); // the new partitions should have been considered as restored state assertTrue(consumerFunction.getRestoredState() != null); assertTrue(!consumerFunction.getSubscribedPartitionsToStartOffsets().isEmpty()); for (Map.Entry<KafkaTopicPartition, Long> expectedEntry : expectedSubscribedPartitionsWithStartOffsets.entrySet()) { assertEquals(expectedEntry.getValue(), consumerFunction.getRestoredState().get(expectedEntry.getKey())); } consumerOperator.close(); consumerOperator.cancel(); }
Example 10
Source File: FlinkKafkaConsumerBaseMigrationTest.java From flink with Apache License 2.0 | 4 votes |
private void writeSnapshot(String path, HashMap<KafkaTopicPartition, Long> state) throws Exception { final OneShotLatch latch = new OneShotLatch(); final AbstractFetcher<String, ?> fetcher = mock(AbstractFetcher.class); doAnswer(new Answer<Void>() { @Override public Void answer(InvocationOnMock invocation) throws Throwable { latch.trigger(); return null; } }).when(fetcher).runFetchLoop(); when(fetcher.snapshotCurrentState()).thenReturn(state); final List<KafkaTopicPartition> partitions = new ArrayList<>(PARTITION_STATE.keySet()); final DummyFlinkKafkaConsumer<String> consumerFunction = new DummyFlinkKafkaConsumer<>(fetcher, TOPICS, partitions, FlinkKafkaConsumerBase.PARTITION_DISCOVERY_DISABLED); StreamSource<String, DummyFlinkKafkaConsumer<String>> consumerOperator = new StreamSource<>(consumerFunction); final AbstractStreamOperatorTestHarness<String> testHarness = new AbstractStreamOperatorTestHarness<>(consumerOperator, 1, 1, 0); testHarness.setTimeCharacteristic(TimeCharacteristic.ProcessingTime); testHarness.setup(); testHarness.open(); final Throwable[] error = new Throwable[1]; // run the source asynchronously Thread runner = new Thread() { @Override public void run() { try { consumerFunction.run(new DummySourceContext() { @Override public void collect(String element) { } }); } catch (Throwable t) { t.printStackTrace(); error[0] = t; } } }; runner.start(); if (!latch.isTriggered()) { latch.await(); } final OperatorSubtaskState snapshot; synchronized (testHarness.getCheckpointLock()) { snapshot = testHarness.snapshot(0L, 0L); } OperatorSnapshotUtil.writeStateHandle(snapshot, path); consumerOperator.close(); runner.join(); }
Example 11
Source File: FromElementsFunctionTest.java From Flink-CEPplus with Apache License 2.0 | 4 votes |
@Test public void testCheckpointAndRestore() { try { final int numElements = 10000; List<Integer> data = new ArrayList<Integer>(numElements); List<Integer> result = new ArrayList<Integer>(numElements); for (int i = 0; i < numElements; i++) { data.add(i); } final FromElementsFunction<Integer> source = new FromElementsFunction<>(IntSerializer.INSTANCE, data); StreamSource<Integer, FromElementsFunction<Integer>> src = new StreamSource<>(source); AbstractStreamOperatorTestHarness<Integer> testHarness = new AbstractStreamOperatorTestHarness<>(src, 1, 1, 0); testHarness.open(); final SourceFunction.SourceContext<Integer> ctx = new ListSourceContext<Integer>(result, 2L); final Throwable[] error = new Throwable[1]; // run the source asynchronously Thread runner = new Thread() { @Override public void run() { try { source.run(ctx); } catch (Throwable t) { error[0] = t; } } }; runner.start(); // wait for a bit Thread.sleep(1000); // make a checkpoint List<Integer> checkpointData = new ArrayList<>(numElements); OperatorSubtaskState handles = null; synchronized (ctx.getCheckpointLock()) { handles = testHarness.snapshot(566, System.currentTimeMillis()); checkpointData.addAll(result); } // cancel the source source.cancel(); runner.join(); // check for errors if (error[0] != null) { System.err.println("Error in asynchronous source runner"); error[0].printStackTrace(); fail("Error in asynchronous source runner"); } final FromElementsFunction<Integer> sourceCopy = new FromElementsFunction<>(IntSerializer.INSTANCE, data); StreamSource<Integer, FromElementsFunction<Integer>> srcCopy = new StreamSource<>(sourceCopy); AbstractStreamOperatorTestHarness<Integer> testHarnessCopy = new AbstractStreamOperatorTestHarness<>(srcCopy, 1, 1, 0); testHarnessCopy.setup(); testHarnessCopy.initializeState(handles); testHarnessCopy.open(); // recovery run SourceFunction.SourceContext<Integer> newCtx = new ListSourceContext<>(checkpointData); sourceCopy.run(newCtx); assertEquals(data, checkpointData); } catch (Exception e) { e.printStackTrace(); fail(e.getMessage()); } }
Example 12
Source File: FlinkKafkaConsumerBaseMigrationTest.java From flink with Apache License 2.0 | 4 votes |
/** * Test restoring from an empty state taken using a previous Flink version, when some partitions could be * found for topics. */ @Test public void testRestoreFromEmptyStateWithPartitions() throws Exception { final List<KafkaTopicPartition> partitions = new ArrayList<>(PARTITION_STATE.keySet()); final DummyFlinkKafkaConsumer<String> consumerFunction = new DummyFlinkKafkaConsumer<>(TOPICS, partitions, FlinkKafkaConsumerBase.PARTITION_DISCOVERY_DISABLED); StreamSource<String, DummyFlinkKafkaConsumer<String>> consumerOperator = new StreamSource<>(consumerFunction); final AbstractStreamOperatorTestHarness<String> testHarness = new AbstractStreamOperatorTestHarness<>(consumerOperator, 1, 1, 0); testHarness.setTimeCharacteristic(TimeCharacteristic.ProcessingTime); testHarness.setup(); // restore state from binary snapshot file testHarness.initializeState( OperatorSnapshotUtil.getResourceFilename( "kafka-consumer-migration-test-flink" + testMigrateVersion + "-empty-state-snapshot")); testHarness.open(); // the expected state in "kafka-consumer-migration-test-flink1.x-snapshot-empty-state"; // all new partitions after the snapshot are considered as partitions that were created while the // consumer wasn't running, and should start from the earliest offset. final HashMap<KafkaTopicPartition, Long> expectedSubscribedPartitionsWithStartOffsets = new HashMap<>(); for (KafkaTopicPartition partition : PARTITION_STATE.keySet()) { expectedSubscribedPartitionsWithStartOffsets.put(partition, KafkaTopicPartitionStateSentinel.EARLIEST_OFFSET); } // assert that there are partitions and is identical to expected list assertTrue(consumerFunction.getSubscribedPartitionsToStartOffsets() != null); assertTrue(!consumerFunction.getSubscribedPartitionsToStartOffsets().isEmpty()); assertEquals(expectedSubscribedPartitionsWithStartOffsets, consumerFunction.getSubscribedPartitionsToStartOffsets()); // the new partitions should have been considered as restored state assertTrue(consumerFunction.getRestoredState() != null); assertTrue(!consumerFunction.getSubscribedPartitionsToStartOffsets().isEmpty()); for (Map.Entry<KafkaTopicPartition, Long> expectedEntry : expectedSubscribedPartitionsWithStartOffsets.entrySet()) { assertEquals(expectedEntry.getValue(), consumerFunction.getRestoredState().get(expectedEntry.getKey())); } consumerOperator.close(); consumerOperator.cancel(); }
Example 13
Source File: FlinkKafkaConsumerBaseMigrationTest.java From Flink-CEPplus with Apache License 2.0 | 4 votes |
private void writeSnapshot(String path, HashMap<KafkaTopicPartition, Long> state) throws Exception { final OneShotLatch latch = new OneShotLatch(); final AbstractFetcher<String, ?> fetcher = mock(AbstractFetcher.class); doAnswer(new Answer<Void>() { @Override public Void answer(InvocationOnMock invocation) throws Throwable { latch.trigger(); return null; } }).when(fetcher).runFetchLoop(); when(fetcher.snapshotCurrentState()).thenReturn(state); final List<KafkaTopicPartition> partitions = new ArrayList<>(PARTITION_STATE.keySet()); final DummyFlinkKafkaConsumer<String> consumerFunction = new DummyFlinkKafkaConsumer<>(fetcher, TOPICS, partitions, FlinkKafkaConsumerBase.PARTITION_DISCOVERY_DISABLED); StreamSource<String, DummyFlinkKafkaConsumer<String>> consumerOperator = new StreamSource<>(consumerFunction); final AbstractStreamOperatorTestHarness<String> testHarness = new AbstractStreamOperatorTestHarness<>(consumerOperator, 1, 1, 0); testHarness.setTimeCharacteristic(TimeCharacteristic.ProcessingTime); testHarness.setup(); testHarness.open(); final Throwable[] error = new Throwable[1]; // run the source asynchronously Thread runner = new Thread() { @Override public void run() { try { consumerFunction.run(new DummySourceContext() { @Override public void collect(String element) { } }); } catch (Throwable t) { t.printStackTrace(); error[0] = t; } } }; runner.start(); if (!latch.isTriggered()) { latch.await(); } final OperatorSubtaskState snapshot; synchronized (testHarness.getCheckpointLock()) { snapshot = testHarness.snapshot(0L, 0L); } OperatorSnapshotUtil.writeStateHandle(snapshot, path); consumerOperator.close(); runner.join(); }
Example 14
Source File: FlinkKinesisConsumerMigrationTest.java From flink with Apache License 2.0 | 4 votes |
@SuppressWarnings("unchecked") private void writeSnapshot(String path, HashMap<StreamShardMetadata, SequenceNumber> state) throws Exception { final List<StreamShardHandle> initialDiscoveryShards = new ArrayList<>(state.size()); for (StreamShardMetadata shardMetadata : state.keySet()) { Shard shard = new Shard(); shard.setShardId(shardMetadata.getShardId()); SequenceNumberRange sequenceNumberRange = new SequenceNumberRange(); sequenceNumberRange.withStartingSequenceNumber("1"); shard.setSequenceNumberRange(sequenceNumberRange); initialDiscoveryShards.add(new StreamShardHandle(shardMetadata.getStreamName(), shard)); } final TestFetcher<String> fetcher = new TestFetcher<>( Collections.singletonList(TEST_STREAM_NAME), new TestSourceContext<>(), new TestRuntimeContext(true, 1, 0), TestUtils.getStandardProperties(), new KinesisDeserializationSchemaWrapper<>(new SimpleStringSchema()), state, initialDiscoveryShards); final DummyFlinkKinesisConsumer<String> consumer = new DummyFlinkKinesisConsumer<>( fetcher, new KinesisDeserializationSchemaWrapper<>(new SimpleStringSchema())); StreamSource<String, DummyFlinkKinesisConsumer<String>> consumerOperator = new StreamSource<>(consumer); final AbstractStreamOperatorTestHarness<String> testHarness = new AbstractStreamOperatorTestHarness<>(consumerOperator, 1, 1, 0); testHarness.setTimeCharacteristic(TimeCharacteristic.ProcessingTime); testHarness.setup(); testHarness.open(); final AtomicReference<Throwable> error = new AtomicReference<>(); // run the source asynchronously Thread runner = new Thread() { @Override public void run() { try { consumer.run(new TestSourceContext<>()); } catch (Throwable t) { t.printStackTrace(); error.set(t); } } }; runner.start(); fetcher.waitUntilRun(); final OperatorSubtaskState snapshot; synchronized (testHarness.getCheckpointLock()) { snapshot = testHarness.snapshot(0L, 0L); } OperatorSnapshotUtil.writeStateHandle(snapshot, path); consumerOperator.close(); runner.join(); }
Example 15
Source File: FlinkKinesisConsumerMigrationTest.java From Flink-CEPplus with Apache License 2.0 | 4 votes |
@Test public void testRestoreWithReshardedStream() throws Exception { final List<StreamShardHandle> initialDiscoveryShards = new ArrayList<>(TEST_STATE.size()); for (StreamShardMetadata shardMetadata : TEST_STATE.keySet()) { // setup the closed shard Shard closedShard = new Shard(); closedShard.setShardId(shardMetadata.getShardId()); SequenceNumberRange closedSequenceNumberRange = new SequenceNumberRange(); closedSequenceNumberRange.withStartingSequenceNumber("1"); closedSequenceNumberRange.withEndingSequenceNumber("1087654321"); // this represents a closed shard closedShard.setSequenceNumberRange(closedSequenceNumberRange); initialDiscoveryShards.add(new StreamShardHandle(shardMetadata.getStreamName(), closedShard)); // setup the new shards Shard newSplitShard1 = new Shard(); newSplitShard1.setShardId(KinesisShardIdGenerator.generateFromShardOrder(1)); SequenceNumberRange newSequenceNumberRange1 = new SequenceNumberRange(); newSequenceNumberRange1.withStartingSequenceNumber("1087654322"); newSplitShard1.setSequenceNumberRange(newSequenceNumberRange1); newSplitShard1.setParentShardId(TEST_SHARD_ID); Shard newSplitShard2 = new Shard(); newSplitShard2.setShardId(KinesisShardIdGenerator.generateFromShardOrder(2)); SequenceNumberRange newSequenceNumberRange2 = new SequenceNumberRange(); newSequenceNumberRange2.withStartingSequenceNumber("2087654322"); newSplitShard2.setSequenceNumberRange(newSequenceNumberRange2); newSplitShard2.setParentShardId(TEST_SHARD_ID); initialDiscoveryShards.add(new StreamShardHandle(shardMetadata.getStreamName(), newSplitShard1)); initialDiscoveryShards.add(new StreamShardHandle(shardMetadata.getStreamName(), newSplitShard2)); } final TestFetcher<String> fetcher = new TestFetcher<>( Collections.singletonList(TEST_STREAM_NAME), new TestSourceContext<>(), new TestRuntimeContext(true, 1, 0), TestUtils.getStandardProperties(), new KinesisDeserializationSchemaWrapper<>(new SimpleStringSchema()), null, initialDiscoveryShards); final DummyFlinkKinesisConsumer<String> consumerFunction = new DummyFlinkKinesisConsumer<>( fetcher, new KinesisDeserializationSchemaWrapper<>(new SimpleStringSchema())); StreamSource<String, DummyFlinkKinesisConsumer<String>> consumerOperator = new StreamSource<>(consumerFunction); final AbstractStreamOperatorTestHarness<String> testHarness = new AbstractStreamOperatorTestHarness<>(consumerOperator, 1, 1, 0); testHarness.setup(); testHarness.initializeState( OperatorSnapshotUtil.getResourceFilename( "kinesis-consumer-migration-test-flink" + testMigrateVersion + "-snapshot")); testHarness.open(); consumerFunction.run(new TestSourceContext<>()); // assert that state is correctly restored assertNotEquals(null, consumerFunction.getRestoredState()); assertEquals(1, consumerFunction.getRestoredState().size()); assertEquals(TEST_STATE, removeEquivalenceWrappers(consumerFunction.getRestoredState())); // assert that the fetcher is registered with all shards, including new shards assertEquals(3, fetcher.getSubscribedShardsState().size()); KinesisStreamShardState restoredClosedShardState = fetcher.getSubscribedShardsState().get(0); assertEquals(TEST_STREAM_NAME, restoredClosedShardState.getStreamShardHandle().getStreamName()); assertEquals(TEST_SHARD_ID, restoredClosedShardState.getStreamShardHandle().getShard().getShardId()); assertTrue(restoredClosedShardState.getStreamShardHandle().isClosed()); assertEquals(TEST_SEQUENCE_NUMBER, restoredClosedShardState.getLastProcessedSequenceNum()); KinesisStreamShardState restoredNewSplitShard1 = fetcher.getSubscribedShardsState().get(1); assertEquals(TEST_STREAM_NAME, restoredNewSplitShard1.getStreamShardHandle().getStreamName()); assertEquals(KinesisShardIdGenerator.generateFromShardOrder(1), restoredNewSplitShard1.getStreamShardHandle().getShard().getShardId()); assertFalse(restoredNewSplitShard1.getStreamShardHandle().isClosed()); // new shards should be consumed from the beginning assertEquals(SentinelSequenceNumber.SENTINEL_EARLIEST_SEQUENCE_NUM.get(), restoredNewSplitShard1.getLastProcessedSequenceNum()); KinesisStreamShardState restoredNewSplitShard2 = fetcher.getSubscribedShardsState().get(2); assertEquals(TEST_STREAM_NAME, restoredNewSplitShard2.getStreamShardHandle().getStreamName()); assertEquals(KinesisShardIdGenerator.generateFromShardOrder(2), restoredNewSplitShard2.getStreamShardHandle().getShard().getShardId()); assertFalse(restoredNewSplitShard2.getStreamShardHandle().isClosed()); // new shards should be consumed from the beginning assertEquals(SentinelSequenceNumber.SENTINEL_EARLIEST_SEQUENCE_NUM.get(), restoredNewSplitShard2.getLastProcessedSequenceNum()); consumerOperator.close(); consumerOperator.cancel(); }
Example 16
Source File: RMQSourceTest.java From flink with Apache License 2.0 | 4 votes |
@Test public void testCheckpointing() throws Exception { source.autoAck = false; StreamSource<String, RMQSource<String>> src = new StreamSource<>(source); AbstractStreamOperatorTestHarness<String> testHarness = new AbstractStreamOperatorTestHarness<>(src, 1, 1, 0); testHarness.open(); sourceThread.start(); Thread.sleep(5); final Random random = new Random(System.currentTimeMillis()); int numSnapshots = 50; long previousSnapshotId; long lastSnapshotId = 0; long totalNumberOfAcks = 0; for (int i = 0; i < numSnapshots; i++) { long snapshotId = random.nextLong(); OperatorSubtaskState data; synchronized (DummySourceContext.lock) { data = testHarness.snapshot(snapshotId, System.currentTimeMillis()); previousSnapshotId = lastSnapshotId; lastSnapshotId = messageId; } // let some time pass Thread.sleep(5); // check if the correct number of messages have been snapshotted final long numIds = lastSnapshotId - previousSnapshotId; RMQTestSource sourceCopy = new RMQTestSource(); StreamSource<String, RMQTestSource> srcCopy = new StreamSource<>(sourceCopy); AbstractStreamOperatorTestHarness<String> testHarnessCopy = new AbstractStreamOperatorTestHarness<>(srcCopy, 1, 1, 0); testHarnessCopy.setup(); testHarnessCopy.initializeState(data); testHarnessCopy.open(); ArrayDeque<Tuple2<Long, Set<String>>> deque = sourceCopy.getRestoredState(); Set<String> messageIds = deque.getLast().f1; assertEquals(numIds, messageIds.size()); if (messageIds.size() > 0) { assertTrue(messageIds.contains(Long.toString(lastSnapshotId))); } // check if the messages are being acknowledged and the transaction committed synchronized (DummySourceContext.lock) { source.notifyCheckpointComplete(snapshotId); } totalNumberOfAcks += numIds; } Mockito.verify(source.channel, Mockito.times((int) totalNumberOfAcks)).basicAck(Mockito.anyLong(), Mockito.eq(false)); Mockito.verify(source.channel, Mockito.times(numSnapshots)).txCommit(); }
Example 17
Source File: FlinkKinesisConsumerMigrationTest.java From flink with Apache License 2.0 | 4 votes |
@Test public void testRestoreWithEmptyState() throws Exception { final List<StreamShardHandle> initialDiscoveryShards = new ArrayList<>(TEST_STATE.size()); for (StreamShardMetadata shardMetadata : TEST_STATE.keySet()) { Shard shard = new Shard(); shard.setShardId(shardMetadata.getShardId()); SequenceNumberRange sequenceNumberRange = new SequenceNumberRange(); sequenceNumberRange.withStartingSequenceNumber("1"); shard.setSequenceNumberRange(sequenceNumberRange); initialDiscoveryShards.add(new StreamShardHandle(shardMetadata.getStreamName(), shard)); } final TestFetcher<String> fetcher = new TestFetcher<>( Collections.singletonList(TEST_STREAM_NAME), new TestSourceContext<>(), new TestRuntimeContext(true, 1, 0), TestUtils.getStandardProperties(), new KinesisDeserializationSchemaWrapper<>(new SimpleStringSchema()), null, initialDiscoveryShards); final DummyFlinkKinesisConsumer<String> consumerFunction = new DummyFlinkKinesisConsumer<>( fetcher, new KinesisDeserializationSchemaWrapper<>(new SimpleStringSchema())); StreamSource<String, DummyFlinkKinesisConsumer<String>> consumerOperator = new StreamSource<>(consumerFunction); final AbstractStreamOperatorTestHarness<String> testHarness = new AbstractStreamOperatorTestHarness<>(consumerOperator, 1, 1, 0); testHarness.setup(); testHarness.initializeState( OperatorSnapshotUtil.getResourceFilename( "kinesis-consumer-migration-test-flink" + testMigrateVersion + "-empty-snapshot")); testHarness.open(); consumerFunction.run(new TestSourceContext<>()); // assert that no state was restored assertTrue(consumerFunction.getRestoredState().isEmpty()); // although the restore state is empty, the fetcher should still have been registered the initial discovered shard; // furthermore, the discovered shard should be considered a newly created shard while the job wasn't running, // and therefore should be consumed from the earliest sequence number KinesisStreamShardState restoredShardState = fetcher.getSubscribedShardsState().get(0); assertEquals(TEST_STREAM_NAME, restoredShardState.getStreamShardHandle().getStreamName()); assertEquals(TEST_SHARD_ID, restoredShardState.getStreamShardHandle().getShard().getShardId()); assertFalse(restoredShardState.getStreamShardHandle().isClosed()); assertEquals(SentinelSequenceNumber.SENTINEL_EARLIEST_SEQUENCE_NUM.get(), restoredShardState.getLastProcessedSequenceNum()); consumerOperator.close(); consumerOperator.cancel(); }
Example 18
Source File: FlinkKafkaConsumerBaseMigrationTest.java From Flink-CEPplus with Apache License 2.0 | 4 votes |
/** * Test restoring from an empty state taken using a previous Flink version, when some partitions could be * found for topics. */ @Test public void testRestoreFromEmptyStateWithPartitions() throws Exception { final List<KafkaTopicPartition> partitions = new ArrayList<>(PARTITION_STATE.keySet()); final DummyFlinkKafkaConsumer<String> consumerFunction = new DummyFlinkKafkaConsumer<>(TOPICS, partitions, FlinkKafkaConsumerBase.PARTITION_DISCOVERY_DISABLED); StreamSource<String, DummyFlinkKafkaConsumer<String>> consumerOperator = new StreamSource<>(consumerFunction); final AbstractStreamOperatorTestHarness<String> testHarness = new AbstractStreamOperatorTestHarness<>(consumerOperator, 1, 1, 0); testHarness.setTimeCharacteristic(TimeCharacteristic.ProcessingTime); testHarness.setup(); // restore state from binary snapshot file testHarness.initializeState( OperatorSnapshotUtil.getResourceFilename( "kafka-consumer-migration-test-flink" + testMigrateVersion + "-empty-state-snapshot")); testHarness.open(); // the expected state in "kafka-consumer-migration-test-flink1.2-snapshot-empty-state"; // all new partitions after the snapshot are considered as partitions that were created while the // consumer wasn't running, and should start from the earliest offset. final HashMap<KafkaTopicPartition, Long> expectedSubscribedPartitionsWithStartOffsets = new HashMap<>(); for (KafkaTopicPartition partition : PARTITION_STATE.keySet()) { expectedSubscribedPartitionsWithStartOffsets.put(partition, KafkaTopicPartitionStateSentinel.EARLIEST_OFFSET); } // assert that there are partitions and is identical to expected list assertTrue(consumerFunction.getSubscribedPartitionsToStartOffsets() != null); assertTrue(!consumerFunction.getSubscribedPartitionsToStartOffsets().isEmpty()); assertEquals(expectedSubscribedPartitionsWithStartOffsets, consumerFunction.getSubscribedPartitionsToStartOffsets()); // the new partitions should have been considered as restored state assertTrue(consumerFunction.getRestoredState() != null); assertTrue(!consumerFunction.getSubscribedPartitionsToStartOffsets().isEmpty()); for (Map.Entry<KafkaTopicPartition, Long> expectedEntry : expectedSubscribedPartitionsWithStartOffsets.entrySet()) { assertEquals(expectedEntry.getValue(), consumerFunction.getRestoredState().get(expectedEntry.getKey())); } consumerOperator.close(); consumerOperator.cancel(); }
Example 19
Source File: ContinuousFileProcessingMigrationTest.java From flink with Apache License 2.0 | 4 votes |
/** * Manually run this to write binary snapshot data. Remove @Ignore to run. */ @Ignore @Test public void writeMonitoringSourceSnapshot() throws Exception { File testFolder = tempFolder.newFolder(); long fileModTime = Long.MIN_VALUE; for (int i = 0; i < 1; i++) { Tuple2<File, String> file = createFileAndFillWithData(testFolder, "file", i, "This is test line."); fileModTime = file.f0.lastModified(); } TextInputFormat format = new TextInputFormat(new Path(testFolder.getAbsolutePath())); final ContinuousFileMonitoringFunction<String> monitoringFunction = new ContinuousFileMonitoringFunction<>(format, FileProcessingMode.PROCESS_CONTINUOUSLY, 1, INTERVAL); StreamSource<TimestampedFileInputSplit, ContinuousFileMonitoringFunction<String>> src = new StreamSource<>(monitoringFunction); final AbstractStreamOperatorTestHarness<TimestampedFileInputSplit> testHarness = new AbstractStreamOperatorTestHarness<>(src, 1, 1, 0); testHarness.open(); final Throwable[] error = new Throwable[1]; final OneShotLatch latch = new OneShotLatch(); // run the source asynchronously Thread runner = new Thread() { @Override public void run() { try { monitoringFunction.run(new DummySourceContext() { @Override public void collect(TimestampedFileInputSplit element) { latch.trigger(); } @Override public void markAsTemporarilyIdle() { } }); } catch (Throwable t) { t.printStackTrace(); error[0] = t; } } }; runner.start(); if (!latch.isTriggered()) { latch.await(); } final OperatorSubtaskState snapshot; synchronized (testHarness.getCheckpointLock()) { snapshot = testHarness.snapshot(0L, 0L); } OperatorSnapshotUtil.writeStateHandle( snapshot, "src/test/resources/monitoring-function-migration-test-" + fileModTime + "-flink" + flinkGenerateSavepointVersion + "-snapshot"); monitoringFunction.cancel(); runner.join(); testHarness.close(); }
Example 20
Source File: ContinuousFileProcessingMigrationTest.java From flink with Apache License 2.0 | 3 votes |
@Test public void testMonitoringSourceRestore() throws Exception { File testFolder = tempFolder.newFolder(); TextInputFormat format = new TextInputFormat(new Path(testFolder.getAbsolutePath())); final ContinuousFileMonitoringFunction<String> monitoringFunction = new ContinuousFileMonitoringFunction<>(format, FileProcessingMode.PROCESS_CONTINUOUSLY, 1, INTERVAL); StreamSource<TimestampedFileInputSplit, ContinuousFileMonitoringFunction<String>> src = new StreamSource<>(monitoringFunction); final AbstractStreamOperatorTestHarness<TimestampedFileInputSplit> testHarness = new AbstractStreamOperatorTestHarness<>(src, 1, 1, 0); testHarness.setup(); testHarness.initializeState( OperatorSnapshotUtil.getResourceFilename( "monitoring-function-migration-test-" + expectedModTime + "-flink" + testMigrateVersion + "-snapshot")); testHarness.open(); Assert.assertEquals((long) expectedModTime, monitoringFunction.getGlobalModificationTime()); }