Search in sources :

Example 81 with DataNodeId

use of com.github.ambry.clustermap.DataNodeId in project ambry by linkedin.

the class ServerPlaintextTest method endToEndCloudBackupTest.

/**
 * Do end to end cloud backup test
 */
@Test
public void endToEndCloudBackupTest() throws Exception {
    assumeTrue(testEncryption);
    plaintextCluster.startServers();
    DataNodeId dataNode = plaintextCluster.getClusterMap().getDataNodeIds().get(0);
    // Start Helix Controller and ZK Server.
    int zkPort = 31999;
    String zkConnectString = "localhost:" + zkPort;
    String vcrClusterName = "vcrTestClusterPlainText";
    TestUtils.ZkInfo zkInfo = new TestUtils.ZkInfo(TestUtils.getTempDir("helixVcr"), "DC1", (byte) 1, zkPort, true);
    ServerTestUtil.endToEndCloudBackupTest(plaintextCluster, zkConnectString, vcrClusterName, dataNode, null, null, notificationSystem, null, false);
    ServerTestUtil.endToEndCloudBackupTest(plaintextCluster, zkConnectString, vcrClusterName, dataNode, null, null, notificationSystem, null, true);
    zkInfo.shutdown();
}
Also used : TestUtils(com.github.ambry.utils.TestUtils) DataNodeId(com.github.ambry.clustermap.DataNodeId) Test(org.junit.Test)

Example 82 with DataNodeId

use of com.github.ambry.clustermap.DataNodeId in project ambry by linkedin.

the class ServerSSLTest method endToEndSSLTest.

@Test
public void endToEndSSLTest() throws IOException, InstantiationException {
    sslCluster.startServers();
    DataNodeId dataNodeId = sslCluster.getGeneralDataNode();
    ServerTestUtil.endToEndTest(new Port(dataNodeId.getSSLPort(), PortType.SSL), "DC1", sslCluster, clientSSLConfig1, clientSSLSocketFactory1, routerProps, testEncryption);
}
Also used : Port(com.github.ambry.network.Port) DataNodeId(com.github.ambry.clustermap.DataNodeId) Test(org.junit.Test)

Example 83 with DataNodeId

use of com.github.ambry.clustermap.DataNodeId in project ambry by linkedin.

the class StatsManagerTest method testReplicaFromOfflineToDropped.

/**
 * Test Offline-To-Dropped transition (both failure and success cases)
 * @throws Exception
 */
@Test
public void testReplicaFromOfflineToDropped() throws Exception {
    ClusterMapConfig clusterMapConfig = new ClusterMapConfig(verifiableProperties);
    ReplicationConfig replicationConfig = new ReplicationConfig(verifiableProperties);
    StoreConfig storeConfig = new StoreConfig(verifiableProperties);
    MockClusterMap clusterMap = new MockClusterMap();
    DataNodeId currentNode = clusterMap.getDataNodeIds().get(0);
    List<ReplicaId> localReplicas = clusterMap.getReplicaIds(currentNode);
    StorageManager storageManager = new StorageManager(storeConfig, new DiskManagerConfig(verifiableProperties), Utils.newScheduler(1, true), new MetricRegistry(), null, clusterMap, currentNode, null, Collections.singletonList(clusterParticipant), new MockTime(), null, new InMemAccountService(false, false));
    storageManager.start();
    MockStoreKeyConverterFactory storeKeyConverterFactory = new MockStoreKeyConverterFactory(null, null);
    storeKeyConverterFactory.setConversionMap(new HashMap<>());
    MockReplicationManager mockReplicationManager = new MockReplicationManager(replicationConfig, clusterMapConfig, storeConfig, storageManager, clusterMap, currentNode, storeKeyConverterFactory, clusterParticipant);
    MockStatsManager mockStatsManager = new MockStatsManager(storageManager, localReplicas, new MetricRegistry(), statsManagerConfig, clusterParticipant);
    // 1. attempt to remove replica while store is still running (remove store failure case)
    ReplicaId replicaToDrop = localReplicas.get(0);
    try {
        clusterParticipant.onPartitionBecomeDroppedFromOffline(replicaToDrop.getPartitionId().toPathString());
        fail("should fail because store is still running");
    } catch (StateTransitionException e) {
        assertEquals("Error code doesn't match", ReplicaOperationFailure, e.getErrorCode());
    }
    // 2. shutdown the store but introduce file deletion failure (put a invalid dir in store dir)
    storageManager.shutdownBlobStore(replicaToDrop.getPartitionId());
    File invalidDir = new File(replicaToDrop.getReplicaPath(), "invalidDir");
    invalidDir.deleteOnExit();
    assertTrue("Couldn't create dir within store dir", invalidDir.mkdir());
    assertTrue("Could not make unreadable", invalidDir.setReadable(false));
    try {
        clusterParticipant.onPartitionBecomeDroppedFromOffline(replicaToDrop.getPartitionId().toPathString());
        fail("should fail because store deletion fails");
    } catch (StateTransitionException e) {
        assertEquals("Error code doesn't match", ReplicaOperationFailure, e.getErrorCode());
    }
    // reset permission to allow deletion to succeed.
    assertTrue("Could not make readable", invalidDir.setReadable(true));
    assertTrue("Could not delete invalid dir", invalidDir.delete());
    // 3. success case (remove another replica because previous replica has been removed from in-mem data structures)
    ReplicaId replica = localReplicas.get(1);
    storageManager.shutdownBlobStore(replica.getPartitionId());
    MockHelixParticipant mockHelixParticipant = Mockito.spy(clusterParticipant);
    doNothing().when(mockHelixParticipant).setPartitionDisabledState(anyString(), anyBoolean());
    mockHelixParticipant.onPartitionBecomeDroppedFromOffline(replica.getPartitionId().toPathString());
    // verify that the replica is no longer present in StorageManager
    assertNull("Store of removed replica should not exist", storageManager.getStore(replica.getPartitionId(), true));
    // purposely remove the same replica in ReplicationManager again to verify it no longer exists
    assertFalse("Should return false because replica no longer exists", mockReplicationManager.removeReplica(replica));
    // purposely remove the same replica in StatsManager again to verify it no longer exists
    assertFalse("Should return false because replica no longer exists", mockStatsManager.removeReplica(replica));
    verify(mockHelixParticipant).setPartitionDisabledState(replica.getPartitionId().toPathString(), false);
    storageManager.shutdown();
    mockStatsManager.shutdown();
}
Also used : DiskManagerConfig(com.github.ambry.config.DiskManagerConfig) MockStoreKeyConverterFactory(com.github.ambry.store.MockStoreKeyConverterFactory) ReplicationConfig(com.github.ambry.config.ReplicationConfig) MetricRegistry(com.codahale.metrics.MetricRegistry) StorageManager(com.github.ambry.store.StorageManager) ClusterMapConfig(com.github.ambry.config.ClusterMapConfig) ReplicaId(com.github.ambry.clustermap.ReplicaId) MockReplicationManager(com.github.ambry.replication.MockReplicationManager) InMemAccountService(com.github.ambry.account.InMemAccountService) MockHelixParticipant(com.github.ambry.clustermap.MockHelixParticipant) StoreConfig(com.github.ambry.config.StoreConfig) DataNodeId(com.github.ambry.clustermap.DataNodeId) MockDataNodeId(com.github.ambry.clustermap.MockDataNodeId) File(java.io.File) MockTime(com.github.ambry.utils.MockTime) MockClusterMap(com.github.ambry.clustermap.MockClusterMap) StateTransitionException(com.github.ambry.clustermap.StateTransitionException) Test(org.junit.Test)

Example 84 with DataNodeId

use of com.github.ambry.clustermap.DataNodeId in project ambry by linkedin.

the class StatsManagerTest method testStatsManagerWithProblematicStores.

/**
 * Test to verify the behavior when dealing with {@link Store} that is null and when {@link StoreException} is thrown.
 * @throws Exception
 */
@Test
public void testStatsManagerWithProblematicStores() throws Exception {
    DataNodeId dataNodeId = new MockDataNodeId(Collections.singletonList(new Port(6667, PortType.PLAINTEXT)), Collections.singletonList("/tmp"), "DC1");
    Map<PartitionId, Store> problematicStoreMap = new HashMap<>();
    PartitionId partitionId1 = new MockPartitionId(1, MockClusterMap.DEFAULT_PARTITION_CLASS, Collections.singletonList((MockDataNodeId) dataNodeId), 0);
    PartitionId partitionId2 = new MockPartitionId(2, MockClusterMap.DEFAULT_PARTITION_CLASS, Collections.singletonList((MockDataNodeId) dataNodeId), 0);
    problematicStoreMap.put(partitionId1, null);
    Store exceptionStore = new MockStore(new MockStoreStats(new HashMap<>(), true));
    problematicStoreMap.put(partitionId2, exceptionStore);
    StatsManager testStatsManager = new StatsManager(new MockStorageManager(problematicStoreMap, dataNodeId), Arrays.asList(partitionId1.getReplicaIds().get(0), partitionId2.getReplicaIds().get(0)), new MetricRegistry(), statsManagerConfig, new MockTime(), null, null, inMemoryAccountService);
    List<PartitionId> unreachablePartitions = new ArrayList<>();
    Map<Long, Map<Short, Map<Short, ContainerStorageStats>>> hostAccountStorageStatsMap = new HashMap<>();
    for (PartitionId partitionId : problematicStoreMap.keySet()) {
        testStatsManager.collectAndAggregateAccountStorageStats(hostAccountStorageStatsMap, partitionId, unreachablePartitions);
    }
    assertEquals("Aggregated map should not contain any value", 0L, hostAccountStorageStatsMap.size());
    assertEquals("Unreachable store count mismatch with expected value", 2, unreachablePartitions.size());
    StatsManager.AccountStatsPublisher publisher = testStatsManager.new AccountStatsPublisher(accountStatsStore);
    publisher.run();
    HostAccountStorageStatsWrapper statsWrapper = accountStatsStore.queryHostAccountStorageStatsByHost("localhost", 0);
    List<String> unreachableStores = statsWrapper.getHeader().getUnreachableStores();
    assertTrue("The unreachable store list should contain Partition1 and Partition2", unreachableStores.containsAll(Arrays.asList(partitionId1.toPathString(), partitionId2.toPathString())));
    // test for the scenario where some stores are healthy and some are bad
    Map<PartitionId, Store> mixedStoreMap = new HashMap<>(storeMap);
    unreachablePartitions.clear();
    PartitionId partitionId3 = new MockPartitionId(3, MockClusterMap.DEFAULT_PARTITION_CLASS, Collections.singletonList((MockDataNodeId) dataNodeId), 0);
    PartitionId partitionId4 = new MockPartitionId(4, MockClusterMap.DEFAULT_PARTITION_CLASS, Collections.singletonList((MockDataNodeId) dataNodeId), 0);
    mixedStoreMap.put(partitionId3, null);
    mixedStoreMap.put(partitionId4, exceptionStore);
    testStatsManager = new StatsManager(new MockStorageManager(mixedStoreMap, dataNodeId), Arrays.asList(partitionId3.getReplicaIds().get(0), partitionId4.getReplicaIds().get(0)), new MetricRegistry(), statsManagerConfig, new MockTime(), null, null, inMemoryAccountService);
    hostAccountStorageStatsMap.clear();
    for (PartitionId partitionId : mixedStoreMap.keySet()) {
        testStatsManager.collectAndAggregateAccountStorageStats(hostAccountStorageStatsMap, partitionId, unreachablePartitions);
    }
    assertEquals("Unreachable store count mismatch with expected value", 2, unreachablePartitions.size());
    // test fetchSnapshot method in StatsManager
    unreachablePartitions.clear();
    // partition 0, 1, 2 are healthy stores, partition 3, 4 are bad ones.
    for (PartitionId partitionId : mixedStoreMap.keySet()) {
        Map<Short, Map<Short, ContainerStorageStats>> containerStatsMapForPartition = hostAccountStorageStatsMap.get(partitionId.getId());
        if (partitionId.getId() < 3) {
            assertEquals("Actual map does not match with expected snapshot with partition id " + partitionId.toPathString(), hostAccountStorageStats.getStorageStats().get(partitionId.getId()), containerStatsMapForPartition);
        }
    }
}
Also used : HashMap(java.util.HashMap) Port(com.github.ambry.network.Port) ArrayList(java.util.ArrayList) InmemoryAccountStatsStore(com.github.ambry.accountstats.InmemoryAccountStatsStore) Store(com.github.ambry.store.Store) ContainerStorageStats(com.github.ambry.server.storagestats.ContainerStorageStats) MockTime(com.github.ambry.utils.MockTime) MockPartitionId(com.github.ambry.clustermap.MockPartitionId) MetricRegistry(com.codahale.metrics.MetricRegistry) MockPartitionId(com.github.ambry.clustermap.MockPartitionId) PartitionId(com.github.ambry.clustermap.PartitionId) MockDataNodeId(com.github.ambry.clustermap.MockDataNodeId) DataNodeId(com.github.ambry.clustermap.DataNodeId) MockDataNodeId(com.github.ambry.clustermap.MockDataNodeId) Map(java.util.Map) HashMap(java.util.HashMap) MockClusterMap(com.github.ambry.clustermap.MockClusterMap) Test(org.junit.Test)

Example 85 with DataNodeId

use of com.github.ambry.clustermap.DataNodeId in project ambry by linkedin.

the class StatsManagerTest method testAddAndRemoveReplica.

/**
 * Test to verify the {@link StatsManager} behaves correctly when dynamically adding/removing {@link ReplicaId}.
 * @throws Exception
 */
@Test
public void testAddAndRemoveReplica() throws Exception {
    // setup testing environment
    Map<PartitionId, Store> testStoreMap = new HashMap<>();
    List<ReplicaId> testReplicas = new ArrayList<>();
    DataNodeId dataNodeId = new MockDataNodeId(Collections.singletonList(new Port(6667, PortType.PLAINTEXT)), Collections.singletonList("/tmp"), "DC1");
    for (int i = 0; i < 3; i++) {
        PartitionId partitionId = new MockPartitionId(i, MockClusterMap.DEFAULT_PARTITION_CLASS, Collections.singletonList((MockDataNodeId) dataNodeId), 0);
        testStoreMap.put(partitionId, new MockStore(new MockStoreStats(hostAccountStorageStats.getStorageStats().get(i), false)));
        testReplicas.add(partitionId.getReplicaIds().get(0));
    }
    StorageManager mockStorageManager = new MockStorageManager(testStoreMap, dataNodeId);
    StatsManager testStatsManager = new StatsManager(mockStorageManager, testReplicas, new MetricRegistry(), statsManagerConfig, new MockTime(), null, null, inMemoryAccountService);
    // verify that adding an existing store to StatsManager should fail
    assertFalse("Adding a store which already exists should fail", testStatsManager.addReplica(testReplicas.get(0)));
    PartitionId partitionId3 = new MockPartitionId(3, MockClusterMap.DEFAULT_PARTITION_CLASS, Collections.singletonList((MockDataNodeId) dataNodeId), 0);
    testStoreMap.put(partitionId3, new MockStore(new MockStoreStats(hostAccountStorageStats.getStorageStats().get(0), false)));
    // verify that partitionId3 is not in stats report before adding to statsManager
    StatsManager.AccountStatsPublisher publisher = testStatsManager.new AccountStatsPublisher(accountStatsStore);
    publisher.run();
    HostAccountStorageStatsWrapper statsWrapper = accountStatsStore.queryHostAccountStorageStatsByHost("localhost", 0);
    assertFalse("Partition3 should not present in stats report", statsWrapper.getStats().getStorageStats().containsKey(partitionId3.getId()));
    // verify that after adding into statsManager, PartitionId3 is in stats report
    testStatsManager.addReplica(partitionId3.getReplicaIds().get(0));
    publisher.run();
    statsWrapper = accountStatsStore.queryHostAccountStorageStatsByHost("localhost", 0);
    assertTrue("Partition3 should present in stats report", statsWrapper.getStats().getStorageStats().containsKey(partitionId3.getId()));
    // verify that after removing PartitionId0 (corresponding to the first replica in replicas list), PartitionId0 is not in the stats report
    PartitionId partitionId0 = testReplicas.get(0).getPartitionId();
    assertTrue("Partition0 should present in stats report before removal", statsWrapper.getStats().getStorageStats().containsKey(partitionId0.getId()));
    testStoreMap.remove(testReplicas.get(0).getPartitionId());
    testStatsManager.removeReplica(testReplicas.get(0));
    publisher.run();
    statsWrapper = accountStatsStore.queryHostAccountStorageStatsByHost("localhost", 0);
    assertFalse("Partition0 should not present in stats report after removal", statsWrapper.getStats().getStorageStats().containsKey(partitionId0.getId()));
    // verify that removing the PartitionId0 should fail because it no longer exists in StatsManager
    assertFalse(testStatsManager.removeReplica(testReplicas.get(0)));
    // concurrent remove test
    CountDownLatch getStatsCountdown1 = new CountDownLatch(1);
    CountDownLatch waitRemoveCountdown = new CountDownLatch(1);
    ((MockStorageManager) mockStorageManager).waitOperationCountdown = waitRemoveCountdown;
    ((MockStorageManager) mockStorageManager).firstCall = true;
    ((MockStorageManager) mockStorageManager).unreachablePartitions.clear();
    for (Store store : testStoreMap.values()) {
        ((MockStore) store).getStatsCountdown = getStatsCountdown1;
        ((MockStore) store).isCollected = false;
    }
    List<PartitionId> partitionRemoved = new ArrayList<>();
    Utils.newThread(() -> {
        // wait until at least one store has been collected (this ensures stats aggregation using old snapshot of map)
        try {
            getStatsCountdown1.await();
        } catch (InterruptedException e) {
            throw new IllegalStateException("CountDown await was interrupted", e);
        }
        // find one store which hasn't been collected
        ReplicaId replicaToRemove = null;
        for (Map.Entry<PartitionId, Store> partitionToStore : testStoreMap.entrySet()) {
            MockStore store = (MockStore) partitionToStore.getValue();
            if (!store.isCollected) {
                replicaToRemove = partitionToStore.getKey().getReplicaIds().get(0);
                break;
            }
        }
        if (replicaToRemove != null) {
            testStatsManager.removeReplica(replicaToRemove);
            testStoreMap.remove(replicaToRemove.getPartitionId());
            partitionRemoved.add(replicaToRemove.getPartitionId());
            // count down to allow stats aggregation to proceed
            waitRemoveCountdown.countDown();
        }
    }, false).start();
    publisher.run();
    statsWrapper = accountStatsStore.queryHostAccountStorageStatsByHost("localhost", 0);
    // verify that the removed store is indeed unreachable during stats aggregation
    assertTrue("The removed partition should be unreachable during aggregation", ((MockStorageManager) mockStorageManager).unreachablePartitions.contains(partitionRemoved.get(0)));
    // verify unreachable store list doesn't contain the store which is removed.
    List<String> unreachableStores = statsWrapper.getHeader().getUnreachableStores();
    assertFalse("The removed partition should not present in unreachable list", unreachableStores.contains(partitionRemoved.get(0).toPathString()));
    // concurrent add test
    CountDownLatch getStatsCountdown2 = new CountDownLatch(1);
    CountDownLatch waitAddCountdown = new CountDownLatch(1);
    ((MockStorageManager) mockStorageManager).waitOperationCountdown = waitAddCountdown;
    ((MockStorageManager) mockStorageManager).firstCall = true;
    ((MockStorageManager) mockStorageManager).unreachablePartitions.clear();
    for (Store store : testStoreMap.values()) {
        ((MockStore) store).getStatsCountdown = getStatsCountdown2;
        ((MockStore) store).isCollected = false;
    }
    PartitionId partitionId4 = new MockPartitionId(4, MockClusterMap.DEFAULT_PARTITION_CLASS, Collections.singletonList((MockDataNodeId) dataNodeId), 0);
    Utils.newThread(() -> {
        // wait until at least one store has been collected (this ensures stats aggregation using old snapshot of map)
        try {
            getStatsCountdown2.await();
        } catch (InterruptedException e) {
            throw new IllegalStateException("CountDown await was interrupted", e);
        }
        testStatsManager.addReplica(partitionId4.getReplicaIds().get(0));
        testStoreMap.put(partitionId4, new MockStore(new MockStoreStats(hostAccountStorageStats.getStorageStats().get(0), false)));
        // count down to allow stats aggregation to proceed
        waitAddCountdown.countDown();
    }, false).start();
    publisher.run();
    statsWrapper = accountStatsStore.queryHostAccountStorageStatsByHost("localhost", 0);
    // verify that new added PartitionId4 is not in report for this round of aggregation
    assertFalse("Partition4 should not present in stats report", statsWrapper.getStats().getStorageStats().containsKey(partitionId4.getId()));
    // verify that new added PartitionId4 will be collected for next round of aggregation
    publisher.run();
    statsWrapper = accountStatsStore.queryHostAccountStorageStatsByHost("localhost", 0);
    assertTrue("Partition4 should present in stats report", statsWrapper.getStats().getStorageStats().containsKey(partitionId4.getId()));
}
Also used : HashMap(java.util.HashMap) Port(com.github.ambry.network.Port) ArrayList(java.util.ArrayList) StorageManager(com.github.ambry.store.StorageManager) InmemoryAccountStatsStore(com.github.ambry.accountstats.InmemoryAccountStatsStore) Store(com.github.ambry.store.Store) MockTime(com.github.ambry.utils.MockTime) MockPartitionId(com.github.ambry.clustermap.MockPartitionId) MetricRegistry(com.codahale.metrics.MetricRegistry) MockPartitionId(com.github.ambry.clustermap.MockPartitionId) PartitionId(com.github.ambry.clustermap.PartitionId) CountDownLatch(java.util.concurrent.CountDownLatch) ReplicaId(com.github.ambry.clustermap.ReplicaId) MockDataNodeId(com.github.ambry.clustermap.MockDataNodeId) DataNodeId(com.github.ambry.clustermap.DataNodeId) MockDataNodeId(com.github.ambry.clustermap.MockDataNodeId) Test(org.junit.Test)

Aggregations

DataNodeId (com.github.ambry.clustermap.DataNodeId)92 ArrayList (java.util.ArrayList)45 Test (org.junit.Test)45 HashMap (java.util.HashMap)29 PartitionId (com.github.ambry.clustermap.PartitionId)28 MockDataNodeId (com.github.ambry.clustermap.MockDataNodeId)27 ReplicaId (com.github.ambry.clustermap.ReplicaId)25 MockClusterMap (com.github.ambry.clustermap.MockClusterMap)23 VerifiableProperties (com.github.ambry.config.VerifiableProperties)23 MetricRegistry (com.codahale.metrics.MetricRegistry)22 MockPartitionId (com.github.ambry.clustermap.MockPartitionId)22 List (java.util.List)22 Map (java.util.Map)22 Port (com.github.ambry.network.Port)21 ClusterMap (com.github.ambry.clustermap.ClusterMap)20 ClusterMapConfig (com.github.ambry.config.ClusterMapConfig)19 StoreKeyFactory (com.github.ambry.store.StoreKeyFactory)18 BlobIdFactory (com.github.ambry.commons.BlobIdFactory)17 HashSet (java.util.HashSet)16 Properties (java.util.Properties)16