Examples with SharedReference org.apache.flink.testutils.junit.SharedReference used on opensource projects

Example 1 with SharedReference

use of org.apache.flink.testutils.junit.SharedReference in project flink by apache.

the class BufferTimeoutITCase method testDisablingBufferTimeout.

/**
 * The test verifies that it is possible to disable explicit buffer flushing. It checks that
 * OutputFlasher thread would not be started when the task is running. But this doesn't
 * guarantee that the unfinished buffers can not be flushed by another events.
 */
@Test
public void testDisablingBufferTimeout() throws Exception {
    final StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();
    env.setParallelism(1);
    env.setBufferTimeout(-1);
    final SharedReference<ArrayList<Integer>> results = sharedObjects.add(new ArrayList<>());
    env.addSource(new SourceFunction<Integer>() {

        @Override
        public void run(SourceContext<Integer> ctx) throws Exception {
            ctx.collect(1);
            // just sleep forever
            Thread.sleep(Long.MAX_VALUE);
        }

        @Override
        public void cancel() {
        }
    }).slotSharingGroup("source").addSink(new SinkFunction<Integer>() {

        @Override
        public void invoke(Integer value, Context context) {
            results.get().add(value);
        }
    }).slotSharingGroup("sink");
    final JobClient jobClient = env.executeAsync();
    CommonTestUtils.waitForAllTaskRunning(MINI_CLUSTER_RESOURCE.getMiniCluster(), jobClient.getJobID(), false);
    assertTrue(RecordWriter.DEFAULT_OUTPUT_FLUSH_THREAD_NAME + " thread is unexpectedly running", Thread.getAllStackTraces().keySet().stream().noneMatch(thread -> thread.getName().startsWith(RecordWriter.DEFAULT_OUTPUT_FLUSH_THREAD_NAME)));
}

Example 2 with SharedReference

use of org.apache.flink.testutils.junit.SharedReference in project flink by apache.

the class SavepointITCase method testTriggerSavepointAndResumeWithNoClaim.

@Test
@Ignore("Disabling this test because it regularly fails on AZP. See FLINK-25427.")
public void testTriggerSavepointAndResumeWithNoClaim() throws Exception {
    final int numTaskManagers = 2;
    final int numSlotsPerTaskManager = 2;
    final int parallelism = numTaskManagers * numSlotsPerTaskManager;
    final StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();
    env.setStateBackend(new EmbeddedRocksDBStateBackend(true));
    env.getCheckpointConfig().enableExternalizedCheckpoints(CheckpointConfig.ExternalizedCheckpointCleanup.RETAIN_ON_CANCELLATION);
    env.getCheckpointConfig().setCheckpointStorage(folder.newFolder().toURI());
    env.setParallelism(parallelism);
    final SharedReference<CountDownLatch> counter = sharedObjects.add(new CountDownLatch(10_000));
    env.fromSequence(1, Long.MAX_VALUE).keyBy(i -> i % parallelism).process(new KeyedProcessFunction<Long, Long, Long>() {

        private ListState<Long> last;

        @Override
        public void open(Configuration parameters) {
            // we use list state here to create sst files of a significant size
            // if sst files do not reach certain thresholds they are not stored
            // in files, but as a byte stream in checkpoints metadata
            last = getRuntimeContext().getListState(new ListStateDescriptor<>("last", BasicTypeInfo.LONG_TYPE_INFO));
        }

        @Override
        public void processElement(Long value, KeyedProcessFunction<Long, Long, Long>.Context ctx, Collector<Long> out) throws Exception {
            last.add(value);
            out.collect(value);
        }
    }).addSink(new SinkFunction<Long>() {

        @Override
        public void invoke(Long value) {
            counter.consumeSync(CountDownLatch::countDown);
        }
    }).setParallelism(1);
    final JobGraph jobGraph = env.getStreamGraph().getJobGraph();
    MiniClusterWithClientResource cluster = new MiniClusterWithClientResource(new MiniClusterResourceConfiguration.Builder().setNumberTaskManagers(numTaskManagers).setNumberSlotsPerTaskManager(numSlotsPerTaskManager).build());
    cluster.before();
    try {
        final JobID jobID1 = new JobID();
        jobGraph.setJobID(jobID1);
        cluster.getClusterClient().submitJob(jobGraph).get();
        CommonTestUtils.waitForAllTaskRunning(cluster.getMiniCluster(), jobID1, false);
        // wait for some records to be processed before taking the checkpoint
        counter.get().await();
        final String firstCheckpoint = cluster.getMiniCluster().triggerCheckpoint(jobID1).get();
        cluster.getClusterClient().cancel(jobID1).get();
        jobGraph.setSavepointRestoreSettings(SavepointRestoreSettings.forPath(firstCheckpoint, false, RestoreMode.NO_CLAIM));
        final JobID jobID2 = new JobID();
        jobGraph.setJobID(jobID2);
        cluster.getClusterClient().submitJob(jobGraph).get();
        CommonTestUtils.waitForAllTaskRunning(cluster.getMiniCluster(), jobID2, false);
        String secondCheckpoint = cluster.getMiniCluster().triggerCheckpoint(jobID2).get();
        cluster.getClusterClient().cancel(jobID2).get();
        // delete the checkpoint we restored from
        FileUtils.deleteDirectory(Paths.get(new URI(firstCheckpoint)).getParent().toFile());
        // we should be able to restore from the second checkpoint even though it has been built
        // on top of the first checkpoint
        jobGraph.setSavepointRestoreSettings(SavepointRestoreSettings.forPath(secondCheckpoint, false, RestoreMode.NO_CLAIM));
        final JobID jobID3 = new JobID();
        jobGraph.setJobID(jobID3);
        cluster.getClusterClient().submitJob(jobGraph).get();
        CommonTestUtils.waitForAllTaskRunning(cluster.getMiniCluster(), jobID3, false);
    } finally {
        cluster.after();
    }
}

Example 3 with SharedReference

use of org.apache.flink.testutils.junit.SharedReference in project flink by apache.

the class FileSinkMigrationITCase method test.

@Test
public void test() throws Exception {
    StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();
    SharedReference<Collection<Long>> list = sharedObjects.add(new ArrayList<>());
    int n = 10000;
    env.setParallelism(100);
    env.fromSequence(0, n).map(i -> list.applySync(l -> l.add(i)));
    env.execute();
    assertEquals(n + 1, list.get().size());
    assertEquals(LongStream.rangeClosed(0, n).boxed().collect(Collectors.toList()), list.get().stream().sorted().collect(Collectors.toList()));
}

Example 4 with SharedReference

use of org.apache.flink.testutils.junit.SharedReference in project flink by apache.

the class CommonExecSinkITCase method testFromValuesWatermarkPropagation.

@Test
public void testFromValuesWatermarkPropagation() throws Exception {
    final StreamTableEnvironment tableEnv = StreamTableEnvironment.create(env);
    final SharedReference<List<Long>> watermarks = sharedObjects.add(new ArrayList<>());
    final SinkFunction<RowData> sinkFunction = new SinkFunction<RowData>() {

        @Override
        public void writeWatermark(org.apache.flink.api.common.eventtime.Watermark watermark) {
            addElement(watermarks, watermark.getTimestamp());
        }
    };
    final TableDescriptor sinkDescriptor = TableFactoryHarness.newBuilder().sink(new TableFactoryHarness.SinkBase() {

        @Override
        public DataStreamSinkProvider getSinkRuntimeProvider(DynamicTableSink.Context context) {
            return (providerContext, dataStream) -> dataStream.addSink(sinkFunction);
        }
    }).build();
    final Table source = tableEnv.fromValues(DataTypes.ROW(DataTypes.FIELD("a", DataTypes.INT())), Row.of(1), Row.of(2), Row.of(3));
    source.executeInsert(sinkDescriptor).await();
    assertThat(watermarks.get().size()).isEqualTo(env.getParallelism());
    for (Long watermark : watermarks.get()) {
        assertThat(watermark).isEqualTo(Watermark.MAX_WATERMARK.getTimestamp());
    }
}

Example 5 with SharedReference

use of org.apache.flink.testutils.junit.SharedReference in project flink by apache.

the class CommonExecSinkITCase method testStreamRecordTimestampInserterDataStreamSinkProvider.

@Test
public void testStreamRecordTimestampInserterDataStreamSinkProvider() throws ExecutionException, InterruptedException {
    final StreamTableEnvironment tableEnv = StreamTableEnvironment.create(env);
    final SharedReference<List<Long>> timestamps = sharedObjects.add(new ArrayList<>());
    final List<Row> rows = Arrays.asList(Row.of(1, "foo", Instant.parse("2020-11-10T11:34:56.123Z")), Row.of(2, "foo", Instant.parse("2020-11-10T12:34:56.789Z")), Row.of(3, "foo", Instant.parse("2020-11-11T10:11:22.777Z")), Row.of(4, "foo", Instant.parse("2020-11-11T10:11:23.888Z")));
    final SinkFunction<RowData> sinkFunction = new SinkFunction<RowData>() {

        @Override
        public void invoke(RowData value, Context context) {
            addElement(timestamps, context.timestamp());
        }
    };
    final TableDescriptor sourceDescriptor = TableFactoryHarness.newBuilder().schema(schemaStreamRecordTimestampInserter(true)).source(new TestSource(rows)).sink(new TableFactoryHarness.SinkBase() {

        @Override
        public DataStreamSinkProvider getSinkRuntimeProvider(DynamicTableSink.Context context) {
            return (providerContext, dataStream) -> dataStream.addSink(sinkFunction);
        }
    }).build();
    tableEnv.createTable("T1", sourceDescriptor);
    final String sqlStmt = "INSERT INTO T1 SELECT * FROM T1";
    assertPlan(tableEnv, sqlStmt, true);
    tableEnv.executeSql(sqlStmt).await();
    Collections.sort(timestamps.get());
    assertTimestampResults(timestamps, rows);
}