Search in sources :

Example 41 with CheckpointId

use of org.apache.samza.checkpoint.CheckpointId in project samza by apache.

the class TestTransactionalStateTaskRestoreManager method testGetStoreActionsForLoggedPersistentStore_RestoreFromLocalToNewestIfCheckpointedOffsetInRangeButMaybeCompacted.

/**
 * This can happen if the changelog offset is valid but the checkpoint is older than min compaction lag ms. E.g., when
 * the job/container shut down and restarted after a long time.
 */
@Test
public void testGetStoreActionsForLoggedPersistentStore_RestoreFromLocalToNewestIfCheckpointedOffsetInRangeButMaybeCompacted() {
    TaskModel mockTaskModel = mock(TaskModel.class);
    TaskName taskName = new TaskName("Partition 0");
    when(mockTaskModel.getTaskName()).thenReturn(taskName);
    Partition taskChangelogPartition = new Partition(0);
    when(mockTaskModel.getChangelogPartition()).thenReturn(taskChangelogPartition);
    String store1Name = "store1";
    StorageEngine store1Engine = mock(StorageEngine.class);
    StoreProperties mockStore1Properties = mock(StoreProperties.class);
    when(store1Engine.getStoreProperties()).thenReturn(mockStore1Properties);
    when(mockStore1Properties.isLoggedStore()).thenReturn(true);
    when(mockStore1Properties.isPersistedToDisk()).thenReturn(true);
    Map<String, StorageEngine> mockStoreEngines = ImmutableMap.of(store1Name, store1Engine);
    String changelog1SystemName = "system1";
    String changelog1StreamName = "store1Changelog";
    SystemStream changelog1SystemStream = new SystemStream(changelog1SystemName, changelog1StreamName);
    SystemStreamPartition changelog1SSP = new SystemStreamPartition(changelog1SystemStream, taskChangelogPartition);
    // checkpointed changelog offset is valid
    SystemStreamPartitionMetadata changelog1SSPMetadata = new SystemStreamPartitionMetadata("4", "20", "21");
    Map<String, SystemStream> mockStoreChangelogs = ImmutableMap.of(store1Name, changelog1SystemStream);
    String changelog1CheckpointedOffset = "5";
    // checkpoint timestamp older than default min compaction lag
    CheckpointId checkpointId = CheckpointId.deserialize("0-0");
    KafkaStateCheckpointMarker kafkaStateCheckpointMarker = new KafkaStateCheckpointMarker(changelog1SSP, changelog1CheckpointedOffset);
    Map<String, KafkaStateCheckpointMarker> mockCheckpointedChangelogOffset = new HashMap<String, KafkaStateCheckpointMarker>() {

        {
            put(store1Name, kafkaStateCheckpointMarker);
        }
    };
    Map<SystemStreamPartition, SystemStreamPartitionMetadata> mockCurrentChangelogOffsets = ImmutableMap.of(changelog1SSP, changelog1SSPMetadata);
    SystemAdmins mockSystemAdmins = mock(SystemAdmins.class);
    SystemAdmin mockSystemAdmin = mock(SystemAdmin.class);
    when(mockSystemAdmins.getSystemAdmin(changelog1SSP.getSystem())).thenReturn(mockSystemAdmin);
    StorageManagerUtil mockStorageManagerUtil = mock(StorageManagerUtil.class);
    File mockLoggedStoreBaseDir = mock(File.class);
    File mockNonLoggedStoreBaseDir = mock(File.class);
    Config mockConfig = mock(Config.class);
    Clock mockClock = mock(Clock.class);
    File mockCurrentStoreDir = mock(File.class);
    File mockStoreNewerCheckpointDir = mock(File.class);
    File mockStoreOlderCheckpointDir = mock(File.class);
    String olderCheckpointDirLocalOffset = "3";
    String newerCheckpointDirLocalOffset = "5";
    when(mockStorageManagerUtil.getTaskStoreDir(eq(mockLoggedStoreBaseDir), eq(store1Name), eq(taskName), any())).thenReturn(mockCurrentStoreDir);
    when(mockStorageManagerUtil.getTaskStoreCheckpointDirs(eq(mockLoggedStoreBaseDir), eq(store1Name), eq(taskName), any())).thenReturn(ImmutableList.of(mockStoreNewerCheckpointDir, mockStoreOlderCheckpointDir));
    when(mockStorageManagerUtil.isLoggedStoreValid(eq(store1Name), eq(mockStoreNewerCheckpointDir), any(), eq(mockStoreChangelogs), eq(mockTaskModel), any(), eq(mockStoreEngines))).thenReturn(true);
    when(mockStorageManagerUtil.isLoggedStoreValid(eq(store1Name), eq(mockStoreOlderCheckpointDir), any(), eq(mockStoreChangelogs), eq(mockTaskModel), any(), eq(mockStoreEngines))).thenReturn(true);
    Set<SystemStreamPartition> mockChangelogSSPs = ImmutableSet.of(changelog1SSP);
    when(mockStorageManagerUtil.readOffsetFile(eq(mockStoreNewerCheckpointDir), eq(mockChangelogSSPs), eq(false))).thenReturn(ImmutableMap.of(changelog1SSP, newerCheckpointDirLocalOffset));
    when(mockStorageManagerUtil.readOffsetFile(eq(mockStoreOlderCheckpointDir), eq(mockChangelogSSPs), eq(false))).thenReturn(// less than checkpointed offset (5)
    ImmutableMap.of(changelog1SSP, olderCheckpointDirLocalOffset));
    Mockito.when(mockSystemAdmin.offsetComparator(anyString(), anyString())).thenAnswer((Answer<Integer>) invocation -> {
        String offset1 = (String) invocation.getArguments()[0];
        String offset2 = (String) invocation.getArguments()[1];
        return Long.valueOf(offset1).compareTo(Long.valueOf(offset2));
    });
    StoreActions storeActions = TransactionalStateTaskRestoreManager.getStoreActions(mockTaskModel, mockStoreEngines, mockStoreChangelogs, mockCheckpointedChangelogOffset, checkpointId, mockCurrentChangelogOffsets, mockSystemAdmins, mockStorageManagerUtil, mockLoggedStoreBaseDir, mockNonLoggedStoreBaseDir, mockConfig, mockClock);
    // ensure that the current store dir and older checkpoint dir are marked for deletion
    assertEquals(2, storeActions.storeDirsToDelete.get(store1Name).size());
    assertTrue(storeActions.storeDirsToDelete.get(store1Name).contains(mockCurrentStoreDir));
    assertTrue(storeActions.storeDirsToDelete.get(store1Name).contains(mockStoreOlderCheckpointDir));
    // ensure that newer checkpoint dir is retained
    assertEquals(1, storeActions.storeDirsToRetain.size());
    assertEquals(mockStoreNewerCheckpointDir, storeActions.storeDirsToRetain.get(store1Name));
    // ensure that we mark the store for restore to head (from local checkpoint to current newest)
    assertEquals("5", storeActions.storesToRestore.get(store1Name).startingOffset);
    assertEquals("20", storeActions.storesToRestore.get(store1Name).endingOffset);
}
Also used : ArrayListMultimap(com.google.common.collect.ArrayListMultimap) ListMultimap(com.google.common.collect.ListMultimap) SSPMetadataCache(org.apache.samza.system.SSPMetadataCache) HashMap(java.util.HashMap) TaskModel(org.apache.samza.job.model.TaskModel) SystemStreamPartition(org.apache.samza.system.SystemStreamPartition) Matchers.anyString(org.mockito.Matchers.anyString) FileUtil(org.apache.samza.util.FileUtil) Answer(org.mockito.stubbing.Answer) ImmutableList(com.google.common.collect.ImmutableList) SystemConsumer(org.apache.samza.system.SystemConsumer) Mockito.verifyNoMoreInteractions(org.mockito.Mockito.verifyNoMoreInteractions) SystemStream(org.apache.samza.system.SystemStream) Matchers.eq(org.mockito.Matchers.eq) Map(java.util.Map) Assert.fail(org.junit.Assert.fail) Path(java.nio.file.Path) MapConfig(org.apache.samza.config.MapConfig) ImmutableSet(com.google.common.collect.ImmutableSet) TaskName(org.apache.samza.container.TaskName) ImmutableMap(com.google.common.collect.ImmutableMap) TaskConfig(org.apache.samza.config.TaskConfig) Assert.assertNotNull(org.junit.Assert.assertNotNull) Partition(org.apache.samza.Partition) Set(java.util.Set) Assert.assertTrue(org.junit.Assert.assertTrue) Clock(org.apache.samza.util.Clock) Test(org.junit.Test) Mockito.times(org.mockito.Mockito.times) Mockito.when(org.mockito.Mockito.when) StoreActions(org.apache.samza.storage.TransactionalStateTaskRestoreManager.StoreActions) File(java.io.File) CheckpointId(org.apache.samza.checkpoint.CheckpointId) Mockito.verify(org.mockito.Mockito.verify) Matchers.any(org.mockito.Matchers.any) Mockito(org.mockito.Mockito) TaskMode(org.apache.samza.job.model.TaskMode) Mockito.never(org.mockito.Mockito.never) Assert.assertNull(org.junit.Assert.assertNull) RestoreOffsets(org.apache.samza.storage.TransactionalStateTaskRestoreManager.RestoreOffsets) SystemAdmin(org.apache.samza.system.SystemAdmin) SystemStreamPartitionMetadata(org.apache.samza.system.SystemStreamMetadata.SystemStreamPartitionMetadata) KafkaStateCheckpointMarker(org.apache.samza.checkpoint.kafka.KafkaStateCheckpointMarker) Config(org.apache.samza.config.Config) Collections(java.util.Collections) Assert.assertEquals(org.junit.Assert.assertEquals) SystemAdmins(org.apache.samza.system.SystemAdmins) Mockito.mock(org.mockito.Mockito.mock) HashMap(java.util.HashMap) MapConfig(org.apache.samza.config.MapConfig) TaskConfig(org.apache.samza.config.TaskConfig) Config(org.apache.samza.config.Config) Matchers.anyString(org.mockito.Matchers.anyString) SystemStreamPartitionMetadata(org.apache.samza.system.SystemStreamMetadata.SystemStreamPartitionMetadata) Clock(org.apache.samza.util.Clock) SystemAdmins(org.apache.samza.system.SystemAdmins) SystemStreamPartition(org.apache.samza.system.SystemStreamPartition) Partition(org.apache.samza.Partition) SystemStream(org.apache.samza.system.SystemStream) StoreActions(org.apache.samza.storage.TransactionalStateTaskRestoreManager.StoreActions) TaskName(org.apache.samza.container.TaskName) CheckpointId(org.apache.samza.checkpoint.CheckpointId) SystemAdmin(org.apache.samza.system.SystemAdmin) File(java.io.File) TaskModel(org.apache.samza.job.model.TaskModel) SystemStreamPartition(org.apache.samza.system.SystemStreamPartition) KafkaStateCheckpointMarker(org.apache.samza.checkpoint.kafka.KafkaStateCheckpointMarker) Test(org.junit.Test)

Example 42 with CheckpointId

use of org.apache.samza.checkpoint.CheckpointId in project samza by apache.

the class TestTransactionalStateTaskRestoreManager method testGetStoreActionsForLoggedPersistentStore_NoRestoreButTrimIfUpToDateStoreCheckpoint.

/**
 * Ensure that we do a trim even if the local offset == checkpointed changelog offset, since there may be
 * additional messages in the changelog since the last commit that we need to revert.
 */
@Test
public void testGetStoreActionsForLoggedPersistentStore_NoRestoreButTrimIfUpToDateStoreCheckpoint() {
    TaskModel mockTaskModel = mock(TaskModel.class);
    TaskName taskName = new TaskName("Partition 0");
    when(mockTaskModel.getTaskName()).thenReturn(taskName);
    Partition taskChangelogPartition = new Partition(0);
    when(mockTaskModel.getChangelogPartition()).thenReturn(taskChangelogPartition);
    String store1Name = "store1";
    StorageEngine store1Engine = mock(StorageEngine.class);
    StoreProperties mockStore1Properties = mock(StoreProperties.class);
    when(store1Engine.getStoreProperties()).thenReturn(mockStore1Properties);
    when(mockStore1Properties.isLoggedStore()).thenReturn(true);
    when(mockStore1Properties.isPersistedToDisk()).thenReturn(true);
    Map<String, StorageEngine> mockStoreEngines = ImmutableMap.of(store1Name, store1Engine);
    String changelog1SystemName = "system1";
    String changelog1StreamName = "store1Changelog";
    SystemStream changelog1SystemStream = new SystemStream(changelog1SystemName, changelog1StreamName);
    SystemStreamPartition changelog1SSP = new SystemStreamPartition(changelog1SystemStream, taskChangelogPartition);
    SystemStreamPartitionMetadata changelog1SSPMetadata = new SystemStreamPartitionMetadata("0", "10", "11");
    Map<String, SystemStream> mockStoreChangelogs = ImmutableMap.of(store1Name, changelog1SystemStream);
    String changelog1CheckpointedOffset = "5";
    CheckpointId checkpointId = CheckpointId.create();
    KafkaStateCheckpointMarker kafkaStateCheckpointMarker = new KafkaStateCheckpointMarker(changelog1SSP, changelog1CheckpointedOffset);
    ImmutableMap<String, KafkaStateCheckpointMarker> mockCheckpointedChangelogOffset = ImmutableMap.of(store1Name, kafkaStateCheckpointMarker);
    Map<SystemStreamPartition, SystemStreamPartitionMetadata> mockCurrentChangelogOffsets = ImmutableMap.of(changelog1SSP, changelog1SSPMetadata);
    SystemAdmins mockSystemAdmins = mock(SystemAdmins.class);
    SystemAdmin mockSystemAdmin = mock(SystemAdmin.class);
    when(mockSystemAdmins.getSystemAdmin(changelog1SSP.getSystem())).thenReturn(mockSystemAdmin);
    StorageManagerUtil mockStorageManagerUtil = mock(StorageManagerUtil.class);
    File mockLoggedStoreBaseDir = mock(File.class);
    File mockNonLoggedStoreBaseDir = mock(File.class);
    Config mockConfig = mock(Config.class);
    Clock mockClock = mock(Clock.class);
    File mockCurrentStoreDir = mock(File.class);
    File mockStoreUpToDateCheckpointDir = mock(File.class);
    File mockStoreOlderCheckpointDir = mock(File.class);
    when(mockStorageManagerUtil.getTaskStoreDir(eq(mockLoggedStoreBaseDir), eq(store1Name), eq(taskName), any())).thenReturn(mockCurrentStoreDir);
    when(mockStorageManagerUtil.getTaskStoreCheckpointDirs(eq(mockLoggedStoreBaseDir), eq(store1Name), eq(taskName), any())).thenReturn(ImmutableList.of(mockStoreUpToDateCheckpointDir, mockStoreOlderCheckpointDir));
    when(mockStorageManagerUtil.isLoggedStoreValid(eq(store1Name), eq(mockStoreUpToDateCheckpointDir), any(), eq(mockStoreChangelogs), eq(mockTaskModel), any(), eq(mockStoreEngines))).thenReturn(true);
    when(mockStorageManagerUtil.isLoggedStoreValid(eq(store1Name), eq(mockStoreOlderCheckpointDir), any(), eq(mockStoreChangelogs), eq(mockTaskModel), any(), eq(mockStoreEngines))).thenReturn(true);
    Set<SystemStreamPartition> mockChangelogSSPs = ImmutableSet.of(changelog1SSP);
    when(mockStorageManagerUtil.readOffsetFile(eq(mockStoreUpToDateCheckpointDir), eq(mockChangelogSSPs), eq(false))).thenReturn(ImmutableMap.of(changelog1SSP, changelog1CheckpointedOffset));
    when(mockStorageManagerUtil.readOffsetFile(eq(mockStoreOlderCheckpointDir), eq(mockChangelogSSPs), eq(false))).thenReturn(// less than checkpointed offset (5)
    ImmutableMap.of(changelog1SSP, "3"));
    Mockito.when(mockSystemAdmin.offsetComparator(anyString(), anyString())).thenAnswer((Answer<Integer>) invocation -> {
        String offset1 = (String) invocation.getArguments()[0];
        String offset2 = (String) invocation.getArguments()[1];
        return Long.valueOf(offset1).compareTo(Long.valueOf(offset2));
    });
    StoreActions storeActions = TransactionalStateTaskRestoreManager.getStoreActions(mockTaskModel, mockStoreEngines, mockStoreChangelogs, mockCheckpointedChangelogOffset, checkpointId, mockCurrentChangelogOffsets, mockSystemAdmins, mockStorageManagerUtil, mockLoggedStoreBaseDir, mockNonLoggedStoreBaseDir, mockConfig, mockClock);
    // ensure that both the current dir and older checkpoint dir are marked for deletion
    assertEquals(2, storeActions.storeDirsToDelete.get(store1Name).size());
    assertTrue(storeActions.storeDirsToDelete.get(store1Name).contains(mockCurrentStoreDir));
    assertTrue(storeActions.storeDirsToDelete.get(store1Name).contains(mockStoreOlderCheckpointDir));
    // ensure that the up-to-date store checkpoint is marked for retention
    assertEquals(mockStoreUpToDateCheckpointDir, storeActions.storeDirsToRetain.get(store1Name));
    // ensure that we mark the store for restore even if local offset == checkpointed offset
    // this is required even if there are no messages to restore, since there may be message we need to trim
    assertEquals(changelog1CheckpointedOffset, storeActions.storesToRestore.get(store1Name).startingOffset);
    assertEquals(changelog1CheckpointedOffset, storeActions.storesToRestore.get(store1Name).endingOffset);
}
Also used : ArrayListMultimap(com.google.common.collect.ArrayListMultimap) ListMultimap(com.google.common.collect.ListMultimap) SSPMetadataCache(org.apache.samza.system.SSPMetadataCache) HashMap(java.util.HashMap) TaskModel(org.apache.samza.job.model.TaskModel) SystemStreamPartition(org.apache.samza.system.SystemStreamPartition) Matchers.anyString(org.mockito.Matchers.anyString) FileUtil(org.apache.samza.util.FileUtil) Answer(org.mockito.stubbing.Answer) ImmutableList(com.google.common.collect.ImmutableList) SystemConsumer(org.apache.samza.system.SystemConsumer) Mockito.verifyNoMoreInteractions(org.mockito.Mockito.verifyNoMoreInteractions) SystemStream(org.apache.samza.system.SystemStream) Matchers.eq(org.mockito.Matchers.eq) Map(java.util.Map) Assert.fail(org.junit.Assert.fail) Path(java.nio.file.Path) MapConfig(org.apache.samza.config.MapConfig) ImmutableSet(com.google.common.collect.ImmutableSet) TaskName(org.apache.samza.container.TaskName) ImmutableMap(com.google.common.collect.ImmutableMap) TaskConfig(org.apache.samza.config.TaskConfig) Assert.assertNotNull(org.junit.Assert.assertNotNull) Partition(org.apache.samza.Partition) Set(java.util.Set) Assert.assertTrue(org.junit.Assert.assertTrue) Clock(org.apache.samza.util.Clock) Test(org.junit.Test) Mockito.times(org.mockito.Mockito.times) Mockito.when(org.mockito.Mockito.when) StoreActions(org.apache.samza.storage.TransactionalStateTaskRestoreManager.StoreActions) File(java.io.File) CheckpointId(org.apache.samza.checkpoint.CheckpointId) Mockito.verify(org.mockito.Mockito.verify) Matchers.any(org.mockito.Matchers.any) Mockito(org.mockito.Mockito) TaskMode(org.apache.samza.job.model.TaskMode) Mockito.never(org.mockito.Mockito.never) Assert.assertNull(org.junit.Assert.assertNull) RestoreOffsets(org.apache.samza.storage.TransactionalStateTaskRestoreManager.RestoreOffsets) SystemAdmin(org.apache.samza.system.SystemAdmin) SystemStreamPartitionMetadata(org.apache.samza.system.SystemStreamMetadata.SystemStreamPartitionMetadata) KafkaStateCheckpointMarker(org.apache.samza.checkpoint.kafka.KafkaStateCheckpointMarker) Config(org.apache.samza.config.Config) Collections(java.util.Collections) Assert.assertEquals(org.junit.Assert.assertEquals) SystemAdmins(org.apache.samza.system.SystemAdmins) Mockito.mock(org.mockito.Mockito.mock) MapConfig(org.apache.samza.config.MapConfig) TaskConfig(org.apache.samza.config.TaskConfig) Config(org.apache.samza.config.Config) Matchers.anyString(org.mockito.Matchers.anyString) SystemStreamPartitionMetadata(org.apache.samza.system.SystemStreamMetadata.SystemStreamPartitionMetadata) Clock(org.apache.samza.util.Clock) SystemAdmins(org.apache.samza.system.SystemAdmins) SystemStreamPartition(org.apache.samza.system.SystemStreamPartition) Partition(org.apache.samza.Partition) SystemStream(org.apache.samza.system.SystemStream) StoreActions(org.apache.samza.storage.TransactionalStateTaskRestoreManager.StoreActions) TaskName(org.apache.samza.container.TaskName) CheckpointId(org.apache.samza.checkpoint.CheckpointId) SystemAdmin(org.apache.samza.system.SystemAdmin) File(java.io.File) TaskModel(org.apache.samza.job.model.TaskModel) SystemStreamPartition(org.apache.samza.system.SystemStreamPartition) KafkaStateCheckpointMarker(org.apache.samza.checkpoint.kafka.KafkaStateCheckpointMarker) Test(org.junit.Test)

Example 43 with CheckpointId

use of org.apache.samza.checkpoint.CheckpointId in project samza by apache.

the class TestTransactionalStateTaskRestoreManager method testGetStoreActionsForLoggedNonPersistentStore_FullRestoreIfCheckpointedOffsetNewerThanNewest.

/**
 * This can happen if the changelog topic was manually deleted and recreated, and the checkpointed/local changelog
 * offset is not valid anymore.
 */
@Test
public void testGetStoreActionsForLoggedNonPersistentStore_FullRestoreIfCheckpointedOffsetNewerThanNewest() {
    TaskModel mockTaskModel = mock(TaskModel.class);
    TaskName taskName = new TaskName("Partition 0");
    when(mockTaskModel.getTaskName()).thenReturn(taskName);
    Partition taskChangelogPartition = new Partition(0);
    when(mockTaskModel.getChangelogPartition()).thenReturn(taskChangelogPartition);
    String store1Name = "store1";
    StorageEngine store1Engine = mock(StorageEngine.class);
    StoreProperties mockStore1Properties = mock(StoreProperties.class);
    when(store1Engine.getStoreProperties()).thenReturn(mockStore1Properties);
    when(mockStore1Properties.isLoggedStore()).thenReturn(true);
    // non-persistent store
    when(mockStore1Properties.isPersistedToDisk()).thenReturn(false);
    Map<String, StorageEngine> mockStoreEngines = ImmutableMap.of(store1Name, store1Engine);
    String changelog1SystemName = "system1";
    String changelog1StreamName = "store1Changelog";
    SystemStream changelog1SystemStream = new SystemStream(changelog1SystemName, changelog1StreamName);
    SystemStreamPartition changelog1SSP = new SystemStreamPartition(changelog1SystemStream, taskChangelogPartition);
    // checkpointed changelog offset > newest offset (e.g. changelog topic got changed)
    SystemStreamPartitionMetadata changelog1SSPMetadata = new SystemStreamPartitionMetadata("0", "10", "11");
    Map<String, SystemStream> mockStoreChangelogs = ImmutableMap.of(store1Name, changelog1SystemStream);
    String changelog1CheckpointedOffset = "21";
    CheckpointId checkpointId = CheckpointId.create();
    KafkaStateCheckpointMarker kafkaStateCheckpointMarker = new KafkaStateCheckpointMarker(changelog1SSP, changelog1CheckpointedOffset);
    Map<String, KafkaStateCheckpointMarker> mockCheckpointedChangelogOffset = new HashMap<String, KafkaStateCheckpointMarker>() {

        {
            put(store1Name, kafkaStateCheckpointMarker);
        }
    };
    Map<SystemStreamPartition, SystemStreamPartitionMetadata> mockCurrentChangelogOffsets = ImmutableMap.of(changelog1SSP, changelog1SSPMetadata);
    SystemAdmins mockSystemAdmins = mock(SystemAdmins.class);
    SystemAdmin mockSystemAdmin = mock(SystemAdmin.class);
    when(mockSystemAdmins.getSystemAdmin(changelog1SSP.getSystem())).thenReturn(mockSystemAdmin);
    StorageManagerUtil mockStorageManagerUtil = mock(StorageManagerUtil.class);
    File mockLoggedStoreBaseDir = mock(File.class);
    File mockNonLoggedStoreBaseDir = mock(File.class);
    Config mockConfig = mock(Config.class);
    Clock mockClock = mock(Clock.class);
    Mockito.when(mockSystemAdmin.offsetComparator(anyString(), anyString())).thenAnswer((Answer<Integer>) invocation -> {
        String offset1 = (String) invocation.getArguments()[0];
        String offset2 = (String) invocation.getArguments()[1];
        return Long.valueOf(offset1).compareTo(Long.valueOf(offset2));
    });
    StoreActions storeActions = TransactionalStateTaskRestoreManager.getStoreActions(mockTaskModel, mockStoreEngines, mockStoreChangelogs, mockCheckpointedChangelogOffset, checkpointId, mockCurrentChangelogOffsets, mockSystemAdmins, mockStorageManagerUtil, mockLoggedStoreBaseDir, mockNonLoggedStoreBaseDir, mockConfig, mockClock);
    // ensure that there is nothing to retain or delete
    assertEquals(0, storeActions.storeDirsToDelete.size());
    assertEquals(0, storeActions.storeDirsToRetain.size());
    // ensure that we mark the store for full restore (from current oldest to current newest)
    assertEquals("0", storeActions.storesToRestore.get(store1Name).startingOffset);
    assertEquals("10", storeActions.storesToRestore.get(store1Name).endingOffset);
}
Also used : ArrayListMultimap(com.google.common.collect.ArrayListMultimap) ListMultimap(com.google.common.collect.ListMultimap) SSPMetadataCache(org.apache.samza.system.SSPMetadataCache) HashMap(java.util.HashMap) TaskModel(org.apache.samza.job.model.TaskModel) SystemStreamPartition(org.apache.samza.system.SystemStreamPartition) Matchers.anyString(org.mockito.Matchers.anyString) FileUtil(org.apache.samza.util.FileUtil) Answer(org.mockito.stubbing.Answer) ImmutableList(com.google.common.collect.ImmutableList) SystemConsumer(org.apache.samza.system.SystemConsumer) Mockito.verifyNoMoreInteractions(org.mockito.Mockito.verifyNoMoreInteractions) SystemStream(org.apache.samza.system.SystemStream) Matchers.eq(org.mockito.Matchers.eq) Map(java.util.Map) Assert.fail(org.junit.Assert.fail) Path(java.nio.file.Path) MapConfig(org.apache.samza.config.MapConfig) ImmutableSet(com.google.common.collect.ImmutableSet) TaskName(org.apache.samza.container.TaskName) ImmutableMap(com.google.common.collect.ImmutableMap) TaskConfig(org.apache.samza.config.TaskConfig) Assert.assertNotNull(org.junit.Assert.assertNotNull) Partition(org.apache.samza.Partition) Set(java.util.Set) Assert.assertTrue(org.junit.Assert.assertTrue) Clock(org.apache.samza.util.Clock) Test(org.junit.Test) Mockito.times(org.mockito.Mockito.times) Mockito.when(org.mockito.Mockito.when) StoreActions(org.apache.samza.storage.TransactionalStateTaskRestoreManager.StoreActions) File(java.io.File) CheckpointId(org.apache.samza.checkpoint.CheckpointId) Mockito.verify(org.mockito.Mockito.verify) Matchers.any(org.mockito.Matchers.any) Mockito(org.mockito.Mockito) TaskMode(org.apache.samza.job.model.TaskMode) Mockito.never(org.mockito.Mockito.never) Assert.assertNull(org.junit.Assert.assertNull) RestoreOffsets(org.apache.samza.storage.TransactionalStateTaskRestoreManager.RestoreOffsets) SystemAdmin(org.apache.samza.system.SystemAdmin) SystemStreamPartitionMetadata(org.apache.samza.system.SystemStreamMetadata.SystemStreamPartitionMetadata) KafkaStateCheckpointMarker(org.apache.samza.checkpoint.kafka.KafkaStateCheckpointMarker) Config(org.apache.samza.config.Config) Collections(java.util.Collections) Assert.assertEquals(org.junit.Assert.assertEquals) SystemAdmins(org.apache.samza.system.SystemAdmins) Mockito.mock(org.mockito.Mockito.mock) HashMap(java.util.HashMap) MapConfig(org.apache.samza.config.MapConfig) TaskConfig(org.apache.samza.config.TaskConfig) Config(org.apache.samza.config.Config) Matchers.anyString(org.mockito.Matchers.anyString) SystemStreamPartitionMetadata(org.apache.samza.system.SystemStreamMetadata.SystemStreamPartitionMetadata) Clock(org.apache.samza.util.Clock) SystemAdmins(org.apache.samza.system.SystemAdmins) SystemStreamPartition(org.apache.samza.system.SystemStreamPartition) Partition(org.apache.samza.Partition) SystemStream(org.apache.samza.system.SystemStream) StoreActions(org.apache.samza.storage.TransactionalStateTaskRestoreManager.StoreActions) TaskName(org.apache.samza.container.TaskName) CheckpointId(org.apache.samza.checkpoint.CheckpointId) SystemAdmin(org.apache.samza.system.SystemAdmin) File(java.io.File) TaskModel(org.apache.samza.job.model.TaskModel) SystemStreamPartition(org.apache.samza.system.SystemStreamPartition) KafkaStateCheckpointMarker(org.apache.samza.checkpoint.kafka.KafkaStateCheckpointMarker) Test(org.junit.Test)

Example 44 with CheckpointId

use of org.apache.samza.checkpoint.CheckpointId in project samza by apache.

the class TestTransactionalStateTaskRestoreManager method testGetStoreActionsForLoggedPersistentStore_RetainOneCheckpointIfMultipleCheckpointsWithSameOffset.

/**
 * This can happen if no new messages were written to the store between commits. There may be more than one store
 * checkpoint if container fails during commit after creating a checkpoint but before deleting the old one.
 */
@Test
public void testGetStoreActionsForLoggedPersistentStore_RetainOneCheckpointIfMultipleCheckpointsWithSameOffset() {
    TaskModel mockTaskModel = mock(TaskModel.class);
    TaskName taskName = new TaskName("Partition 0");
    when(mockTaskModel.getTaskName()).thenReturn(taskName);
    Partition taskChangelogPartition = new Partition(0);
    when(mockTaskModel.getChangelogPartition()).thenReturn(taskChangelogPartition);
    String store1Name = "store1";
    StorageEngine store1Engine = mock(StorageEngine.class);
    StoreProperties mockStore1Properties = mock(StoreProperties.class);
    when(store1Engine.getStoreProperties()).thenReturn(mockStore1Properties);
    when(mockStore1Properties.isLoggedStore()).thenReturn(true);
    when(mockStore1Properties.isPersistedToDisk()).thenReturn(true);
    Map<String, StorageEngine> mockStoreEngines = ImmutableMap.of(store1Name, store1Engine);
    String changelog1SystemName = "system1";
    String changelog1StreamName = "store1Changelog";
    SystemStream changelog1SystemStream = new SystemStream(changelog1SystemName, changelog1StreamName);
    SystemStreamPartition changelog1SSP = new SystemStreamPartition(changelog1SystemStream, taskChangelogPartition);
    SystemStreamPartitionMetadata changelog1SSPMetadata = new SystemStreamPartitionMetadata("0", "10", "11");
    Map<String, SystemStream> mockStoreChangelogs = ImmutableMap.of(store1Name, changelog1SystemStream);
    String changelog1CheckpointedOffset = "5";
    CheckpointId checkpointId = CheckpointId.create();
    KafkaStateCheckpointMarker kafkaStateCheckpointMarker = new KafkaStateCheckpointMarker(changelog1SSP, changelog1CheckpointedOffset);
    ImmutableMap<String, KafkaStateCheckpointMarker> mockCheckpointedChangelogOffset = ImmutableMap.of(store1Name, kafkaStateCheckpointMarker);
    Map<SystemStreamPartition, SystemStreamPartitionMetadata> mockCurrentChangelogOffsets = ImmutableMap.of(changelog1SSP, changelog1SSPMetadata);
    SystemAdmins mockSystemAdmins = mock(SystemAdmins.class);
    SystemAdmin mockSystemAdmin = mock(SystemAdmin.class);
    when(mockSystemAdmins.getSystemAdmin(changelog1SSP.getSystem())).thenReturn(mockSystemAdmin);
    StorageManagerUtil mockStorageManagerUtil = mock(StorageManagerUtil.class);
    File mockLoggedStoreBaseDir = mock(File.class);
    File mockNonLoggedStoreBaseDir = mock(File.class);
    Config mockConfig = mock(Config.class);
    Clock mockClock = mock(Clock.class);
    File mockCurrentStoreDir = mock(File.class);
    File mockStoreNewerCheckpointDir = mock(File.class);
    File mockStoreOlderCheckpointDir = mock(File.class);
    when(mockStorageManagerUtil.getTaskStoreDir(eq(mockLoggedStoreBaseDir), eq(store1Name), eq(taskName), any())).thenReturn(mockCurrentStoreDir);
    when(mockStorageManagerUtil.getTaskStoreCheckpointDirs(eq(mockLoggedStoreBaseDir), eq(store1Name), eq(taskName), any())).thenReturn(ImmutableList.of(mockStoreNewerCheckpointDir, mockStoreOlderCheckpointDir));
    when(mockStorageManagerUtil.isLoggedStoreValid(eq(store1Name), eq(mockStoreNewerCheckpointDir), any(), eq(mockStoreChangelogs), eq(mockTaskModel), any(), eq(mockStoreEngines))).thenReturn(true);
    when(mockStorageManagerUtil.isLoggedStoreValid(eq(store1Name), eq(mockStoreOlderCheckpointDir), any(), eq(mockStoreChangelogs), eq(mockTaskModel), any(), eq(mockStoreEngines))).thenReturn(true);
    Set<SystemStreamPartition> mockChangelogSSPs = ImmutableSet.of(changelog1SSP);
    when(mockStorageManagerUtil.readOffsetFile(eq(mockStoreNewerCheckpointDir), eq(mockChangelogSSPs), eq(false))).thenReturn(// equal to checkpointed offset (5)
    ImmutableMap.of(changelog1SSP, changelog1CheckpointedOffset));
    when(mockStorageManagerUtil.readOffsetFile(eq(mockStoreOlderCheckpointDir), eq(mockChangelogSSPs), eq(false))).thenReturn(// also equal to checkpointed offset (5)
    ImmutableMap.of(changelog1SSP, changelog1CheckpointedOffset));
    Mockito.when(mockSystemAdmin.offsetComparator(anyString(), anyString())).thenAnswer((Answer<Integer>) invocation -> {
        String offset1 = (String) invocation.getArguments()[0];
        String offset2 = (String) invocation.getArguments()[1];
        return Long.valueOf(offset1).compareTo(Long.valueOf(offset2));
    });
    StoreActions storeActions = TransactionalStateTaskRestoreManager.getStoreActions(mockTaskModel, mockStoreEngines, mockStoreChangelogs, mockCheckpointedChangelogOffset, checkpointId, mockCurrentChangelogOffsets, mockSystemAdmins, mockStorageManagerUtil, mockLoggedStoreBaseDir, mockNonLoggedStoreBaseDir, mockConfig, mockClock);
    // ensure that both the current dir and one of the checkpoint dirs are marked for deletion
    assertEquals(2, storeActions.storeDirsToDelete.get(store1Name).size());
    assertTrue(storeActions.storeDirsToDelete.get(store1Name).contains(mockCurrentStoreDir));
    // ensure that the one of the store checkpoint is marked for retention
    assertNotNull(storeActions.storeDirsToRetain.get(store1Name));
    // ensure that we mark the store for restore even if local offset == checkpointed offset
    // this is required since there may be message we need to trim
    assertEquals(changelog1CheckpointedOffset, storeActions.storesToRestore.get(store1Name).startingOffset);
    assertEquals(changelog1CheckpointedOffset, storeActions.storesToRestore.get(store1Name).endingOffset);
}
Also used : ArrayListMultimap(com.google.common.collect.ArrayListMultimap) ListMultimap(com.google.common.collect.ListMultimap) SSPMetadataCache(org.apache.samza.system.SSPMetadataCache) HashMap(java.util.HashMap) TaskModel(org.apache.samza.job.model.TaskModel) SystemStreamPartition(org.apache.samza.system.SystemStreamPartition) Matchers.anyString(org.mockito.Matchers.anyString) FileUtil(org.apache.samza.util.FileUtil) Answer(org.mockito.stubbing.Answer) ImmutableList(com.google.common.collect.ImmutableList) SystemConsumer(org.apache.samza.system.SystemConsumer) Mockito.verifyNoMoreInteractions(org.mockito.Mockito.verifyNoMoreInteractions) SystemStream(org.apache.samza.system.SystemStream) Matchers.eq(org.mockito.Matchers.eq) Map(java.util.Map) Assert.fail(org.junit.Assert.fail) Path(java.nio.file.Path) MapConfig(org.apache.samza.config.MapConfig) ImmutableSet(com.google.common.collect.ImmutableSet) TaskName(org.apache.samza.container.TaskName) ImmutableMap(com.google.common.collect.ImmutableMap) TaskConfig(org.apache.samza.config.TaskConfig) Assert.assertNotNull(org.junit.Assert.assertNotNull) Partition(org.apache.samza.Partition) Set(java.util.Set) Assert.assertTrue(org.junit.Assert.assertTrue) Clock(org.apache.samza.util.Clock) Test(org.junit.Test) Mockito.times(org.mockito.Mockito.times) Mockito.when(org.mockito.Mockito.when) StoreActions(org.apache.samza.storage.TransactionalStateTaskRestoreManager.StoreActions) File(java.io.File) CheckpointId(org.apache.samza.checkpoint.CheckpointId) Mockito.verify(org.mockito.Mockito.verify) Matchers.any(org.mockito.Matchers.any) Mockito(org.mockito.Mockito) TaskMode(org.apache.samza.job.model.TaskMode) Mockito.never(org.mockito.Mockito.never) Assert.assertNull(org.junit.Assert.assertNull) RestoreOffsets(org.apache.samza.storage.TransactionalStateTaskRestoreManager.RestoreOffsets) SystemAdmin(org.apache.samza.system.SystemAdmin) SystemStreamPartitionMetadata(org.apache.samza.system.SystemStreamMetadata.SystemStreamPartitionMetadata) KafkaStateCheckpointMarker(org.apache.samza.checkpoint.kafka.KafkaStateCheckpointMarker) Config(org.apache.samza.config.Config) Collections(java.util.Collections) Assert.assertEquals(org.junit.Assert.assertEquals) SystemAdmins(org.apache.samza.system.SystemAdmins) Mockito.mock(org.mockito.Mockito.mock) MapConfig(org.apache.samza.config.MapConfig) TaskConfig(org.apache.samza.config.TaskConfig) Config(org.apache.samza.config.Config) Matchers.anyString(org.mockito.Matchers.anyString) SystemStreamPartitionMetadata(org.apache.samza.system.SystemStreamMetadata.SystemStreamPartitionMetadata) Clock(org.apache.samza.util.Clock) SystemAdmins(org.apache.samza.system.SystemAdmins) SystemStreamPartition(org.apache.samza.system.SystemStreamPartition) Partition(org.apache.samza.Partition) SystemStream(org.apache.samza.system.SystemStream) StoreActions(org.apache.samza.storage.TransactionalStateTaskRestoreManager.StoreActions) TaskName(org.apache.samza.container.TaskName) CheckpointId(org.apache.samza.checkpoint.CheckpointId) SystemAdmin(org.apache.samza.system.SystemAdmin) File(java.io.File) TaskModel(org.apache.samza.job.model.TaskModel) SystemStreamPartition(org.apache.samza.system.SystemStreamPartition) KafkaStateCheckpointMarker(org.apache.samza.checkpoint.kafka.KafkaStateCheckpointMarker) Test(org.junit.Test)

Example 45 with CheckpointId

use of org.apache.samza.checkpoint.CheckpointId in project samza by apache.

the class TestBlobStoreRestoreManager method testRestoreRetainsCheckpointDirsIfValid.

@Test
public void testRestoreRetainsCheckpointDirsIfValid() throws IOException {
    String jobName = "testJobName";
    String jobId = "testJobId";
    TaskName taskName = mock(TaskName.class);
    BlobStoreRestoreManagerMetrics metrics = new BlobStoreRestoreManagerMetrics(new MetricsRegistryMap());
    metrics.initStoreMetrics(ImmutableList.of("storeName"));
    Set<String> storesToRestore = ImmutableSet.of("storeName");
    SnapshotIndex snapshotIndex = mock(SnapshotIndex.class);
    Map<String, Pair<String, SnapshotIndex>> prevStoreSnapshotIndexes = ImmutableMap.of("storeName", Pair.of("blobId", snapshotIndex));
    DirIndex dirIndex = BlobStoreTestUtil.createDirIndex("[a]");
    when(snapshotIndex.getDirIndex()).thenReturn(dirIndex);
    CheckpointId checkpointId = CheckpointId.create();
    when(snapshotIndex.getSnapshotMetadata()).thenReturn(new SnapshotMetadata(checkpointId, "jobName", "jobId", "taskName", "storeName"));
    Path loggedBaseDir = Files.createTempDirectory(BlobStoreTestUtil.TEMP_DIR_PREFIX);
    // create store dir to be deleted during restore
    Path storeDir = Files.createTempDirectory(loggedBaseDir, "storeDir-");
    // create checkpoint dir so that shouldRestore = false (areSameDir == true later)
    Path storeCheckpointDir = Files.createTempDirectory(loggedBaseDir, "storeDir-" + checkpointId + "-");
    // create a dummy file to verify after dir rename.
    Path tempFile = Files.createTempFile(storeCheckpointDir, "tempFile-", null);
    StorageConfig storageConfig = mock(StorageConfig.class);
    StorageManagerUtil storageManagerUtil = mock(StorageManagerUtil.class);
    when(storageManagerUtil.getTaskStoreDir(eq(loggedBaseDir.toFile()), eq("storeName"), eq(taskName), eq(TaskMode.Active))).thenReturn(storeDir.toFile());
    when(storageManagerUtil.getStoreCheckpointDir(any(File.class), eq(checkpointId))).thenReturn(storeCheckpointDir.toString());
    when(storageManagerUtil.getTaskStoreCheckpointDirs(any(File.class), anyString(), any(TaskName.class), any(TaskMode.class))).thenReturn(ImmutableList.of(storeCheckpointDir.toFile()));
    BlobStoreUtil blobStoreUtil = mock(BlobStoreUtil.class);
    DirDiffUtil dirDiffUtil = mock(DirDiffUtil.class);
    // ensures shouldRestore is not called
    when(dirDiffUtil.areSameDir(anySet(), anyBoolean())).thenReturn((arg1, arg2) -> true);
    // return immediately without restoring.
    when(blobStoreUtil.restoreDir(eq(storeDir.toFile()), eq(dirIndex), any(Metadata.class))).thenReturn(CompletableFuture.completedFuture(null));
    BlobStoreRestoreManager.restoreStores(jobName, jobId, taskName, storesToRestore, prevStoreSnapshotIndexes, loggedBaseDir.toFile(), storageConfig, metrics, storageManagerUtil, blobStoreUtil, dirDiffUtil, EXECUTOR);
    // verify that the store directory restore was not called (should have restored from checkpoint dir)
    verify(blobStoreUtil, times(0)).restoreDir(eq(storeDir.toFile()), eq(dirIndex), any(Metadata.class));
    // verify that the checkpoint dir was renamed to store dir
    assertFalse(storeCheckpointDir.toFile().exists());
    assertTrue(storeDir.toFile().exists());
    assertTrue(Files.exists(Paths.get(storeDir.toString(), tempFile.getFileName().toString())));
}
Also used : Path(java.nio.file.Path) SnapshotIndex(org.apache.samza.storage.blobstore.index.SnapshotIndex) StorageConfig(org.apache.samza.config.StorageConfig) StorageManagerUtil(org.apache.samza.storage.StorageManagerUtil) BlobStoreRestoreManagerMetrics(org.apache.samza.storage.blobstore.metrics.BlobStoreRestoreManagerMetrics) SnapshotMetadata(org.apache.samza.storage.blobstore.index.SnapshotMetadata) BlobStoreUtil(org.apache.samza.storage.blobstore.util.BlobStoreUtil) Mockito.anyString(org.mockito.Mockito.anyString) TaskMode(org.apache.samza.job.model.TaskMode) TaskName(org.apache.samza.container.TaskName) SnapshotMetadata(org.apache.samza.storage.blobstore.index.SnapshotMetadata) CheckpointId(org.apache.samza.checkpoint.CheckpointId) DirIndex(org.apache.samza.storage.blobstore.index.DirIndex) MetricsRegistryMap(org.apache.samza.metrics.MetricsRegistryMap) File(java.io.File) DirDiffUtil(org.apache.samza.storage.blobstore.util.DirDiffUtil) Pair(org.apache.commons.lang3.tuple.Pair) Test(org.junit.Test)

Aggregations

CheckpointId (org.apache.samza.checkpoint.CheckpointId)53 Test (org.junit.Test)48 File (java.io.File)44 HashMap (java.util.HashMap)43 Map (java.util.Map)42 Path (java.nio.file.Path)41 ImmutableMap (com.google.common.collect.ImmutableMap)40 TaskName (org.apache.samza.container.TaskName)38 Collections (java.util.Collections)36 Partition (org.apache.samza.Partition)34 SystemStreamPartition (org.apache.samza.system.SystemStreamPartition)34 Answer (org.mockito.stubbing.Answer)34 MapConfig (org.apache.samza.config.MapConfig)31 TaskMode (org.apache.samza.job.model.TaskMode)30 FileUtil (org.apache.samza.util.FileUtil)30 Mockito (org.mockito.Mockito)30 Set (java.util.Set)29 ImmutableList (com.google.common.collect.ImmutableList)28 ImmutableSet (com.google.common.collect.ImmutableSet)28 SystemStream (org.apache.samza.system.SystemStream)27