Example #1
0
        /// <summary>
        /// Test that the NN initializes its under-replicated blocks queue
        /// before it is ready to exit safemode (HDFS-1476)
        /// </summary>
        /// <exception cref="System.Exception"/>
        public virtual void TestInitializeReplQueuesEarly()
        {
            Log.Info("Starting testInitializeReplQueuesEarly");
            // Spray the blocks around the cluster when we add DNs instead of
            // concentrating all blocks on the first node.
            BlockManagerTestUtil.SetWritingPrefersLocalNode(cluster.GetNamesystem().GetBlockManager
                                                                (), false);
            cluster.StartDataNodes(conf, 2, true, HdfsServerConstants.StartupOption.Regular,
                                   null);
            cluster.WaitActive();
            Log.Info("Creating files");
            DFSTestUtil.CreateFile(fs, TestPath, 15 * BlockSize, (short)1, 1L);
            Log.Info("Stopping all DataNodes");
            IList <MiniDFSCluster.DataNodeProperties> dnprops = Lists.NewLinkedList();

            dnprops.AddItem(cluster.StopDataNode(0));
            dnprops.AddItem(cluster.StopDataNode(0));
            dnprops.AddItem(cluster.StopDataNode(0));
            cluster.GetConfiguration(0).SetFloat(DFSConfigKeys.DfsNamenodeReplQueueThresholdPctKey
                                                 , 1f / 15f);
            Log.Info("Restarting NameNode");
            cluster.RestartNameNode();
            NameNode nn     = cluster.GetNameNode();
            string   status = nn.GetNamesystem().GetSafemode();

            NUnit.Framework.Assert.AreEqual("Safe mode is ON. The reported blocks 0 needs additional "
                                            + "15 blocks to reach the threshold 0.9990 of total blocks 15." + Newline + "The number of live datanodes 0 has reached the minimum number 0. "
                                            + "Safe mode will be turned off automatically once the thresholds " + "have been reached."
                                            , status);
            NUnit.Framework.Assert.IsFalse("Mis-replicated block queues should not be initialized "
                                           + "until threshold is crossed", NameNodeAdapter.SafeModeInitializedReplQueues(nn
                                                                                                                         ));
            Log.Info("Restarting one DataNode");
            cluster.RestartDataNode(dnprops.Remove(0));
            // Wait for block reports from all attached storages of
            // the restarted DN to come in.
            GenericTestUtils.WaitFor(new _Supplier_214(this), 10, 10000);
            int safe = NameNodeAdapter.GetSafeModeSafeBlocks(nn);

            NUnit.Framework.Assert.IsTrue("Expected first block report to make some blocks safe."
                                          , safe > 0);
            NUnit.Framework.Assert.IsTrue("Did not expect first block report to make all blocks safe."
                                          , safe < 15);
            NUnit.Framework.Assert.IsTrue(NameNodeAdapter.SafeModeInitializedReplQueues(nn));
            // Ensure that UnderReplicatedBlocks goes up to 15 - safe. Misreplicated
            // blocks are processed asynchronously so this may take a few seconds.
            // Failure here will manifest as a test timeout.
            BlockManagerTestUtil.UpdateState(nn.GetNamesystem().GetBlockManager());
            long underReplicatedBlocks = nn.GetNamesystem().GetUnderReplicatedBlocks();

            while (underReplicatedBlocks != (15 - safe))
            {
                Log.Info("UnderReplicatedBlocks expected=" + (15 - safe) + ", actual=" + underReplicatedBlocks
                         );
                Sharpen.Thread.Sleep(100);
                BlockManagerTestUtil.UpdateState(nn.GetNamesystem().GetBlockManager());
                underReplicatedBlocks = nn.GetNamesystem().GetUnderReplicatedBlocks();
            }
            cluster.RestartDataNodes();
        }
Example #2
0
        public virtual void EnsureSerialNumbersNeverOverlap()
        {
            BlockTokenSecretManager btsm1 = cluster.GetNamesystem(0).GetBlockManager().GetBlockTokenSecretManager
                                                ();
            BlockTokenSecretManager btsm2 = cluster.GetNamesystem(1).GetBlockManager().GetBlockTokenSecretManager
                                                ();

            btsm1.SetSerialNo(0);
            btsm2.SetSerialNo(0);
            NUnit.Framework.Assert.IsFalse(btsm1.GetSerialNoForTesting() == btsm2.GetSerialNoForTesting
                                               ());
            btsm1.SetSerialNo(int.MaxValue);
            btsm2.SetSerialNo(int.MaxValue);
            NUnit.Framework.Assert.IsFalse(btsm1.GetSerialNoForTesting() == btsm2.GetSerialNoForTesting
                                               ());
            btsm1.SetSerialNo(int.MinValue);
            btsm2.SetSerialNo(int.MinValue);
            NUnit.Framework.Assert.IsFalse(btsm1.GetSerialNoForTesting() == btsm2.GetSerialNoForTesting
                                               ());
            btsm1.SetSerialNo(int.MaxValue / 2);
            btsm2.SetSerialNo(int.MaxValue / 2);
            NUnit.Framework.Assert.IsFalse(btsm1.GetSerialNoForTesting() == btsm2.GetSerialNoForTesting
                                               ());
            btsm1.SetSerialNo(int.MinValue / 2);
            btsm2.SetSerialNo(int.MinValue / 2);
            NUnit.Framework.Assert.IsFalse(btsm1.GetSerialNoForTesting() == btsm2.GetSerialNoForTesting
                                               ());
        }
        private static void AssertCTimesEqual(MiniDFSCluster cluster)
        {
            long nn1CTime = cluster.GetNamesystem(0).GetFSImage().GetStorage().GetCTime();
            long nn2CTime = cluster.GetNamesystem(1).GetFSImage().GetStorage().GetCTime();

            NUnit.Framework.Assert.AreEqual(nn1CTime, nn2CTime);
        }
        public virtual void TestConfigureMinValidVolumes()
        {
            Assume.AssumeTrue(!Runtime.GetProperty("os.name").StartsWith("Windows"));
            // Bring up two additional datanodes that need both of their volumes
            // functioning in order to stay up.
            conf.SetInt(DFSConfigKeys.DfsDatanodeFailedVolumesToleratedKey, 0);
            cluster.StartDataNodes(conf, 2, true, null, null);
            cluster.WaitActive();
            DatanodeManager dm = cluster.GetNamesystem().GetBlockManager().GetDatanodeManager
                                     ();
            long origCapacity = DFSTestUtil.GetLiveDatanodeCapacity(dm);
            long dnCapacity   = DFSTestUtil.GetDatanodeCapacity(dm, 0);
            // Fail a volume on the 2nd DN
            FilePath dn2Vol1 = new FilePath(dataDir, "data" + (2 * 1 + 1));

            DataNodeTestUtils.InjectDataDirFailure(dn2Vol1);
            // Should only get two replicas (the first DN and the 3rd)
            Path file1 = new Path("/test1");

            DFSTestUtil.CreateFile(fs, file1, 1024, (short)3, 1L);
            DFSTestUtil.WaitReplication(fs, file1, (short)2);
            // Check that this single failure caused a DN to die.
            DFSTestUtil.WaitForDatanodeStatus(dm, 2, 1, 0, origCapacity - (1 * dnCapacity), WaitForHeartbeats
                                              );
            // If we restore the volume we should still only be able to get
            // two replicas since the DN is still considered dead.
            DataNodeTestUtils.RestoreDataDirFromFailure(dn2Vol1);
            Path file2 = new Path("/test2");

            DFSTestUtil.CreateFile(fs, file2, 1024, (short)3, 1L);
            DFSTestUtil.WaitReplication(fs, file2, (short)2);
        }
        public virtual void TestRetryCacheRebuild()
        {
            DFSTestUtil.RunOperations(cluster, filesystem, conf, BlockSize, 0);
            FSNamesystem namesystem = cluster.GetNamesystem();
            LightWeightCache <RetryCache.CacheEntry, RetryCache.CacheEntry> cacheSet = (LightWeightCache
                                                                                        <RetryCache.CacheEntry, RetryCache.CacheEntry>)namesystem.GetRetryCache().GetCacheSet
                                                                                           ();

            NUnit.Framework.Assert.AreEqual(25, cacheSet.Size());
            IDictionary <RetryCache.CacheEntry, RetryCache.CacheEntry> oldEntries = new Dictionary
                                                                                    <RetryCache.CacheEntry, RetryCache.CacheEntry>();
            IEnumerator <RetryCache.CacheEntry> iter = cacheSet.GetEnumerator();

            while (iter.HasNext())
            {
                RetryCache.CacheEntry entry = iter.Next();
                oldEntries[entry] = entry;
            }
            // restart NameNode
            cluster.RestartNameNode();
            cluster.WaitActive();
            namesystem = cluster.GetNamesystem();
            // check retry cache
            NUnit.Framework.Assert.IsTrue(namesystem.HasRetryCache());
            cacheSet = (LightWeightCache <RetryCache.CacheEntry, RetryCache.CacheEntry>)namesystem
                       .GetRetryCache().GetCacheSet();
            NUnit.Framework.Assert.AreEqual(25, cacheSet.Size());
            iter = cacheSet.GetEnumerator();
            while (iter.HasNext())
            {
                RetryCache.CacheEntry entry = iter.Next();
                NUnit.Framework.Assert.IsTrue(oldEntries.Contains(entry));
            }
        }
        public virtual void TestCorruptBlock()
        {
            // Create a file with single block with two replicas
            Path file = GetTestPath("testCorruptBlock");

            CreateFile(file, 100, (short)2);
            // Corrupt first replica of the block
            LocatedBlock block = NameNodeAdapter.GetBlockLocations(cluster.GetNameNode(), file
                                                                   .ToString(), 0, 1).Get(0);

            cluster.GetNamesystem().WriteLock();
            try
            {
                bm.FindAndMarkBlockAsCorrupt(block.GetBlock(), block.GetLocations()[0], "STORAGE_ID"
                                             , "TEST");
            }
            finally
            {
                cluster.GetNamesystem().WriteUnlock();
            }
            UpdateMetrics();
            MetricsRecordBuilder rb = MetricsAsserts.GetMetrics(NsMetrics);

            MetricsAsserts.AssertGauge("CorruptBlocks", 1L, rb);
            MetricsAsserts.AssertGauge("PendingReplicationBlocks", 1L, rb);
            MetricsAsserts.AssertGauge("ScheduledReplicationBlocks", 1L, rb);
            fs.Delete(file, true);
            rb = WaitForDnMetricValue(NsMetrics, "CorruptBlocks", 0L);
            MetricsAsserts.AssertGauge("PendingReplicationBlocks", 0L, rb);
            MetricsAsserts.AssertGauge("ScheduledReplicationBlocks", 0L, rb);
        }
        public virtual void TestSnapshottableDirs()
        {
            cluster.GetNamesystem().GetSnapshotManager().SetAllowNestedSnapshots(true);
            MetricsAsserts.AssertGauge("SnapshottableDirectories", 0, MetricsAsserts.GetMetrics
                                           (NsMetrics));
            MetricsAsserts.AssertCounter("AllowSnapshotOps", 0L, MetricsAsserts.GetMetrics(NnMetrics
                                                                                           ));
            MetricsAsserts.AssertCounter("DisallowSnapshotOps", 0L, MetricsAsserts.GetMetrics
                                             (NnMetrics));
            // Allow snapshots for directories, and check the metrics
            hdfs.AllowSnapshot(sub1);
            MetricsAsserts.AssertGauge("SnapshottableDirectories", 1, MetricsAsserts.GetMetrics
                                           (NsMetrics));
            MetricsAsserts.AssertCounter("AllowSnapshotOps", 1L, MetricsAsserts.GetMetrics(NnMetrics
                                                                                           ));
            Path sub2 = new Path(dir, "sub2");
            Path file = new Path(sub2, "file");

            DFSTestUtil.CreateFile(hdfs, file, 1024, Replication, seed);
            hdfs.AllowSnapshot(sub2);
            MetricsAsserts.AssertGauge("SnapshottableDirectories", 2, MetricsAsserts.GetMetrics
                                           (NsMetrics));
            MetricsAsserts.AssertCounter("AllowSnapshotOps", 2L, MetricsAsserts.GetMetrics(NnMetrics
                                                                                           ));
            Path subsub1 = new Path(sub1, "sub1sub1");
            Path subfile = new Path(subsub1, "file");

            DFSTestUtil.CreateFile(hdfs, subfile, 1024, Replication, seed);
            hdfs.AllowSnapshot(subsub1);
            MetricsAsserts.AssertGauge("SnapshottableDirectories", 3, MetricsAsserts.GetMetrics
                                           (NsMetrics));
            MetricsAsserts.AssertCounter("AllowSnapshotOps", 3L, MetricsAsserts.GetMetrics(NnMetrics
                                                                                           ));
            // Set an already snapshottable directory to snapshottable, should not
            // change the metrics
            hdfs.AllowSnapshot(sub1);
            MetricsAsserts.AssertGauge("SnapshottableDirectories", 3, MetricsAsserts.GetMetrics
                                           (NsMetrics));
            // But the number of allowSnapshot operations still increases
            MetricsAsserts.AssertCounter("AllowSnapshotOps", 4L, MetricsAsserts.GetMetrics(NnMetrics
                                                                                           ));
            // Disallow the snapshot for snapshottable directories, then check the
            // metrics again
            hdfs.DisallowSnapshot(sub1);
            MetricsAsserts.AssertGauge("SnapshottableDirectories", 2, MetricsAsserts.GetMetrics
                                           (NsMetrics));
            MetricsAsserts.AssertCounter("DisallowSnapshotOps", 1L, MetricsAsserts.GetMetrics
                                             (NnMetrics));
            // delete subsub1, snapshottable directories should be 1
            hdfs.Delete(subsub1, true);
            MetricsAsserts.AssertGauge("SnapshottableDirectories", 1, MetricsAsserts.GetMetrics
                                           (NsMetrics));
            // list all the snapshottable directories
            SnapshottableDirectoryStatus[] status = hdfs.GetSnapshottableDirListing();
            NUnit.Framework.Assert.AreEqual(1, status.Length);
            MetricsAsserts.AssertCounter("ListSnapshottableDirOps", 1L, MetricsAsserts.GetMetrics
                                             (NnMetrics));
        }
Example #8
0
 private static void LowerKeyUpdateIntervalAndClearKeys(MiniDFSCluster cluster)
 {
     LowerKeyUpdateIntervalAndClearKeys(cluster.GetNamesystem(0));
     LowerKeyUpdateIntervalAndClearKeys(cluster.GetNamesystem(1));
     foreach (DataNode dn in cluster.GetDataNodes())
     {
         dn.ClearAllBlockSecretKeys();
     }
 }
Example #9
0
        public virtual void TestAddBlock()
        {
            DistributedFileSystem fs = cluster.GetFileSystem();
            Path file1 = new Path("/file1");
            Path file2 = new Path("/file2");
            Path file3 = new Path("/file3");
            Path file4 = new Path("/file4");

            DFSTestUtil.CreateFile(fs, file1, Blocksize - 1, Replication, 0L);
            DFSTestUtil.CreateFile(fs, file2, Blocksize, Replication, 0L);
            DFSTestUtil.CreateFile(fs, file3, Blocksize * 2 - 1, Replication, 0L);
            DFSTestUtil.CreateFile(fs, file4, Blocksize * 2, Replication, 0L);
            // restart NameNode
            cluster.RestartNameNode(true);
            FSDirectory fsdir = cluster.GetNamesystem().GetFSDirectory();
            // check file1
            INodeFile file1Node = fsdir.GetINode4Write(file1.ToString()).AsFile();

            BlockInfoContiguous[] file1Blocks = file1Node.GetBlocks();
            NUnit.Framework.Assert.AreEqual(1, file1Blocks.Length);
            NUnit.Framework.Assert.AreEqual(Blocksize - 1, file1Blocks[0].GetNumBytes());
            NUnit.Framework.Assert.AreEqual(HdfsServerConstants.BlockUCState.Complete, file1Blocks
                                            [0].GetBlockUCState());
            // check file2
            INodeFile file2Node = fsdir.GetINode4Write(file2.ToString()).AsFile();

            BlockInfoContiguous[] file2Blocks = file2Node.GetBlocks();
            NUnit.Framework.Assert.AreEqual(1, file2Blocks.Length);
            NUnit.Framework.Assert.AreEqual(Blocksize, file2Blocks[0].GetNumBytes());
            NUnit.Framework.Assert.AreEqual(HdfsServerConstants.BlockUCState.Complete, file2Blocks
                                            [0].GetBlockUCState());
            // check file3
            INodeFile file3Node = fsdir.GetINode4Write(file3.ToString()).AsFile();

            BlockInfoContiguous[] file3Blocks = file3Node.GetBlocks();
            NUnit.Framework.Assert.AreEqual(2, file3Blocks.Length);
            NUnit.Framework.Assert.AreEqual(Blocksize, file3Blocks[0].GetNumBytes());
            NUnit.Framework.Assert.AreEqual(HdfsServerConstants.BlockUCState.Complete, file3Blocks
                                            [0].GetBlockUCState());
            NUnit.Framework.Assert.AreEqual(Blocksize - 1, file3Blocks[1].GetNumBytes());
            NUnit.Framework.Assert.AreEqual(HdfsServerConstants.BlockUCState.Complete, file3Blocks
                                            [1].GetBlockUCState());
            // check file4
            INodeFile file4Node = fsdir.GetINode4Write(file4.ToString()).AsFile();

            BlockInfoContiguous[] file4Blocks = file4Node.GetBlocks();
            NUnit.Framework.Assert.AreEqual(2, file4Blocks.Length);
            NUnit.Framework.Assert.AreEqual(Blocksize, file4Blocks[0].GetNumBytes());
            NUnit.Framework.Assert.AreEqual(HdfsServerConstants.BlockUCState.Complete, file4Blocks
                                            [0].GetBlockUCState());
            NUnit.Framework.Assert.AreEqual(Blocksize, file4Blocks[1].GetNumBytes());
            NUnit.Framework.Assert.AreEqual(HdfsServerConstants.BlockUCState.Complete, file4Blocks
                                            [1].GetBlockUCState());
        }
        /// <exception cref="System.IO.IOException"/>
        private LocatedBlocks CreateFileGetBlocks(string filenamePrefix)
        {
            Path filePath = new Path("/" + filenamePrefix + ".dat");

            // Write out a file with a few blocks, get block locations.
            DFSTestUtil.CreateFile(fs, filePath, BlockSize, BlockSize * NumBlocks, BlockSize,
                                   NumDatanodes, seed);
            // Get the block list for the file with the block locations.
            LocatedBlocks blocks = client.GetLocatedBlocks(filePath.ToString(), 0, BlockSize
                                                           * NumBlocks);

            Assert.AssertThat(cluster.GetNamesystem().GetUnderReplicatedBlocks(), IS.Is(0L));
            return(blocks);
        }
        /// <summary>
        /// Run file operations to create edits for all op codes
        /// to be tested.
        /// </summary>
        /// <remarks>
        /// Run file operations to create edits for all op codes
        /// to be tested.
        /// the following op codes are deprecated and therefore not tested:
        /// OP_DATANODE_ADD    ( 5)
        /// OP_DATANODE_REMOVE ( 6)
        /// OP_SET_NS_QUOTA    (11)
        /// OP_CLEAR_NS_QUOTA  (12)
        /// </remarks>
        /// <exception cref="System.IO.IOException"/>
        private CheckpointSignature RunOperations()
        {
            Log.Info("Creating edits by performing fs operations");
            // no check, if it's not it throws an exception which is what we want
            DistributedFileSystem dfs = cluster.GetFileSystem();

            DFSTestUtil.RunOperations(cluster, dfs, cluster.GetConfiguration(0), dfs.GetDefaultBlockSize
                                          (), 0);
            // OP_ROLLING_UPGRADE_START
            cluster.GetNamesystem().GetEditLog().LogStartRollingUpgrade(Time.Now());
            // OP_ROLLING_UPGRADE_FINALIZE
            cluster.GetNamesystem().GetEditLog().LogFinalizeRollingUpgrade(Time.Now());
            // Force a roll so we get an OP_END_LOG_SEGMENT txn
            return(cluster.GetNameNodeRpc().RollEditLog());
        }
Example #12
0
        /// <exception cref="System.Exception"/>
        public virtual void TestQuery()
        {
            Configuration    conf    = new Configuration();
            MiniQJMHACluster cluster = null;

            try
            {
                cluster = new MiniQJMHACluster.Builder(conf).Build();
                MiniDFSCluster dfsCluster = cluster.GetDfsCluster();
                dfsCluster.WaitActive();
                dfsCluster.TransitionToActive(0);
                DistributedFileSystem dfs = dfsCluster.GetFileSystem(0);
                dfsCluster.ShutdownNameNode(1);
                // start rolling upgrade
                RollingUpgradeInfo info = dfs.RollingUpgrade(HdfsConstants.RollingUpgradeAction.Prepare
                                                             );
                NUnit.Framework.Assert.IsTrue(info.IsStarted());
                info = dfs.RollingUpgrade(HdfsConstants.RollingUpgradeAction.Query);
                NUnit.Framework.Assert.IsFalse(info.CreatedRollbackImages());
                dfsCluster.RestartNameNode(1);
                QueryForPreparation(dfs);
                // The NN should have a copy of the fsimage in case of rollbacks.
                NUnit.Framework.Assert.IsTrue(dfsCluster.GetNamesystem(0).GetFSImage().HasRollbackFSImage
                                                  ());
            }
            finally
            {
                if (cluster != null)
                {
                    cluster.Shutdown();
                }
            }
        }
Example #13
0
        public static void SetUp()
        {
            conf = new HdfsConfiguration();
            conf.SetBoolean(DFSConfigKeys.DfsNamenodeReplicationConsiderloadKey, false);
            // Set up the hosts/exclude files.
            localFileSys = FileSystem.GetLocal(conf);
            Path workingDir = localFileSys.GetWorkingDirectory();

            dir = new Path(workingDir, "build/test/data/work-dir/decommission");
            NUnit.Framework.Assert.IsTrue(localFileSys.Mkdirs(dir));
            excludeFile = new Path(dir, "exclude");
            conf.Set(DFSConfigKeys.DfsHostsExclude, excludeFile.ToUri().GetPath());
            Path includeFile = new Path(dir, "include");

            conf.Set(DFSConfigKeys.DfsHosts, includeFile.ToUri().GetPath());
            conf.SetInt(DFSConfigKeys.DfsNamenodeHeartbeatRecheckIntervalKey, 1000);
            conf.SetInt(DFSConfigKeys.DfsHeartbeatIntervalKey, 1);
            conf.SetInt(DFSConfigKeys.DfsNamenodeReplicationPendingTimeoutSecKey, 4);
            conf.SetInt(DFSConfigKeys.DfsNamenodeReplicationIntervalKey, 1000);
            conf.SetInt(DFSConfigKeys.DfsNamenodeDecommissionIntervalKey, 1);
            conf.SetLong(DFSConfigKeys.DfsDatanodeBalanceBandwidthpersecKey, 1);
            WriteConfigFile(localFileSys, excludeFile, null);
            WriteConfigFile(localFileSys, includeFile, null);
            cluster = new MiniDFSCluster.Builder(conf).NumDataNodes(numDatanodes).Build();
            cluster.WaitActive();
            fileSys = cluster.GetFileSystem();
            cluster.GetNamesystem().GetBlockManager().GetDatanodeManager().SetHeartbeatExpireInterval
                (3000);
            Logger.GetLogger(typeof(DecommissionManager)).SetLevel(Level.Debug);
        }
Example #14
0
        public virtual void TestBlocksScheduledCounter()
        {
            cluster = new MiniDFSCluster.Builder(new HdfsConfiguration()).Build();
            cluster.WaitActive();
            fs = cluster.GetFileSystem();
            //open a file an write a few bytes:
            FSDataOutputStream @out = fs.Create(new Path("/testBlockScheduledCounter"));

            for (int i = 0; i < 1024; i++)
            {
                @out.Write(i);
            }
            // flush to make sure a block is allocated.
            @out.Hflush();
            AList <DatanodeDescriptor> dnList = new AList <DatanodeDescriptor>();
            DatanodeManager            dm     = cluster.GetNamesystem().GetBlockManager().GetDatanodeManager
                                                    ();

            dm.FetchDatanodes(dnList, dnList, false);
            DatanodeDescriptor dn = dnList[0];

            NUnit.Framework.Assert.AreEqual(1, dn.GetBlocksScheduled());
            // close the file and the counter should go to zero.
            @out.Close();
            NUnit.Framework.Assert.AreEqual(0, dn.GetBlocksScheduled());
        }
 /// <exception cref="System.IO.IOException"/>
 private void CorruptBlock(MiniDFSCluster cluster, FileSystem fs, Path fileName, int
                           dnIndex, ExtendedBlock block)
 {
     // corrupt the block on datanode dnIndex
     // the indexes change once the nodes are restarted.
     // But the datadirectory will not change
     NUnit.Framework.Assert.IsTrue(cluster.CorruptReplica(dnIndex, block));
     MiniDFSCluster.DataNodeProperties dnProps = cluster.StopDataNode(0);
     // Each datanode has multiple data dirs, check each
     for (int dirIndex = 0; dirIndex < 2; dirIndex++)
     {
         string   bpid        = cluster.GetNamesystem().GetBlockPoolId();
         FilePath storageDir  = cluster.GetStorageDir(dnIndex, dirIndex);
         FilePath dataDir     = MiniDFSCluster.GetFinalizedDir(storageDir, bpid);
         FilePath scanLogFile = new FilePath(dataDir, "dncp_block_verification.log.curr");
         if (scanLogFile.Exists())
         {
             // wait for one minute for deletion to succeed;
             for (int i = 0; !scanLogFile.Delete(); i++)
             {
                 NUnit.Framework.Assert.IsTrue("Could not delete log file in one minute", i < 60);
                 try
                 {
                     Sharpen.Thread.Sleep(1000);
                 }
                 catch (Exception)
                 {
                 }
             }
         }
     }
     // restart the detained so the corrupt replica will be detected
     cluster.RestartDataNode(dnProps);
 }
        /// <exception cref="System.IO.IOException"/>
        private void TestRbwReplicas(MiniDFSCluster cluster, bool isCorrupt)
        {
            FSDataOutputStream @out = null;
            FileSystem         fs   = cluster.GetFileSystem();
            Path src = new Path("/test.txt");

            try
            {
                int fileLen = 515;
                // create some rbw replicas on disk
                byte[] writeBuf = new byte[fileLen];
                new Random().NextBytes(writeBuf);
                @out = fs.Create(src);
                @out.Write(writeBuf);
                @out.Hflush();
                DataNode dn = cluster.GetDataNodes()[0];
                foreach (FsVolumeSpi v in Dataset(dn).GetVolumes())
                {
                    FsVolumeImpl volume     = (FsVolumeImpl)v;
                    FilePath     currentDir = volume.GetCurrentDir().GetParentFile().GetParentFile();
                    FilePath     rbwDir     = new FilePath(currentDir, "rbw");
                    foreach (FilePath file in rbwDir.ListFiles())
                    {
                        if (isCorrupt && Block.IsBlockFilename(file))
                        {
                            new RandomAccessFile(file, "rw").SetLength(fileLen - 1);
                        }
                    }
                }
                // corrupt
                cluster.RestartDataNodes();
                cluster.WaitActive();
                dn = cluster.GetDataNodes()[0];
                // check volumeMap: one rwr replica
                string     bpid     = cluster.GetNamesystem().GetBlockPoolId();
                ReplicaMap replicas = Dataset(dn).volumeMap;
                NUnit.Framework.Assert.AreEqual(1, replicas.Size(bpid));
                ReplicaInfo replica = replicas.Replicas(bpid).GetEnumerator().Next();
                NUnit.Framework.Assert.AreEqual(HdfsServerConstants.ReplicaState.Rwr, replica.GetState
                                                    ());
                if (isCorrupt)
                {
                    NUnit.Framework.Assert.AreEqual((fileLen - 1) / 512 * 512, replica.GetNumBytes());
                }
                else
                {
                    NUnit.Framework.Assert.AreEqual(fileLen, replica.GetNumBytes());
                }
                Dataset(dn).Invalidate(bpid, new Block[] { replica });
            }
            finally
            {
                IOUtils.CloseStream(@out);
                if (fs.Exists(src))
                {
                    fs.Delete(src, false);
                }
                fs.Close();
            }
        }
Example #17
0
        /// <summary>Initializes the cluster.</summary>
        /// <param name="numDataNodes">number of datanodes</param>
        /// <param name="storagesPerDatanode">number of storage locations on each datanode</param>
        /// <param name="failedVolumesTolerated">number of acceptable volume failures</param>
        /// <exception cref="System.Exception">if there is any failure</exception>
        private void InitCluster(int numDataNodes, int storagesPerDatanode, int failedVolumesTolerated
                                 )
        {
            conf = new HdfsConfiguration();
            conf.SetLong(DFSConfigKeys.DfsBlockSizeKey, 512L);

            /*
             * Lower the DN heartbeat, DF rate, and recheck interval to one second
             * so state about failures and datanode death propagates faster.
             */
            conf.SetInt(DFSConfigKeys.DfsHeartbeatIntervalKey, 1);
            conf.SetInt(DFSConfigKeys.DfsDfIntervalKey, 1000);
            conf.SetInt(DFSConfigKeys.DfsNamenodeHeartbeatRecheckIntervalKey, 1000);
            conf.SetInt(DFSConfigKeys.DfsDatanodeFailedVolumesToleratedKey, failedVolumesTolerated
                        );
            cluster = new MiniDFSCluster.Builder(conf).NumDataNodes(numDataNodes).StoragesPerDatanode
                          (storagesPerDatanode).Build();
            cluster.WaitActive();
            fs      = cluster.GetFileSystem();
            dataDir = cluster.GetDataDirectory();
            long dnCapacity = DFSTestUtil.GetDatanodeCapacity(cluster.GetNamesystem().GetBlockManager
                                                                  ().GetDatanodeManager(), 0);

            volumeCapacity = dnCapacity / cluster.GetStoragesPerDatanode();
        }
Example #18
0
        public virtual void Pipeline_01()
        {
            string MethodName = GenericTestUtils.GetMethodName();

            if (Log.IsDebugEnabled())
            {
                Log.Debug("Running " + MethodName);
            }
            Path filePath = new Path("/" + MethodName + ".dat");

            DFSTestUtil.CreateFile(fs, filePath, FileSize, ReplFactor, rand.NextLong());
            if (Log.IsDebugEnabled())
            {
                Log.Debug("Invoking append but doing nothing otherwise...");
            }
            FSDataOutputStream ofs = fs.Append(filePath);

            ofs.WriteBytes("Some more stuff to write");
            ((DFSOutputStream)ofs.GetWrappedStream()).Hflush();
            IList <LocatedBlock> lb = cluster.GetNameNodeRpc().GetBlockLocations(filePath.ToString
                                                                                     (), FileSize - 1, FileSize).GetLocatedBlocks();
            string bpid = cluster.GetNamesystem().GetBlockPoolId();

            foreach (DataNode dn in cluster.GetDataNodes())
            {
                Replica r = DataNodeTestUtils.FetchReplicaInfo(dn, bpid, lb[0].GetBlock().GetBlockId
                                                                   ());
                NUnit.Framework.Assert.IsTrue("Replica on DN " + dn + " shouldn't be null", r !=
                                              null);
                NUnit.Framework.Assert.AreEqual("Should be RBW replica on " + dn + " after sequence of calls append()/write()/hflush()"
                                                , HdfsServerConstants.ReplicaState.Rbw, r.GetState());
            }
            ofs.Close();
        }
Example #19
0
 /// <exception cref="System.Exception"/>
 internal TestContext(Configuration conf, int numNameServices)
 {
     this.numNameServices = numNameServices;
     MiniDFSCluster.Builder bld = new MiniDFSCluster.Builder(conf).NumDataNodes(1).StoragesPerDatanode
                                      (1);
     if (numNameServices > 1)
     {
         bld.NnTopology(MiniDFSNNTopology.SimpleFederatedTopology(numNameServices));
     }
     cluster = bld.Build();
     cluster.WaitActive();
     dfs = new DistributedFileSystem[numNameServices];
     for (int i = 0; i < numNameServices; i++)
     {
         dfs[i] = cluster.GetFileSystem(i);
     }
     bpids = new string[numNameServices];
     for (int i_1 = 0; i_1 < numNameServices; i_1++)
     {
         bpids[i_1] = cluster.GetNamesystem(i_1).GetBlockPoolId();
     }
     datanode     = cluster.GetDataNodes()[0];
     blockScanner = datanode.GetBlockScanner();
     for (int i_2 = 0; i_2 < numNameServices; i_2++)
     {
         dfs[i_2].Mkdirs(new Path("/test"));
     }
     data    = datanode.GetFSDataset();
     volumes = data.GetVolumes();
 }
        /// <summary>Pull out one of the JournalAndStream objects from the edit log.</summary>
        private JournalSet.JournalAndStream GetJournalAndStream(int index)
        {
            FSImage   fsimage = cluster.GetNamesystem().GetFSImage();
            FSEditLog editLog = fsimage.GetEditLog();

            return(editLog.GetJournals()[index]);
        }
 public virtual void SetUp()
 {
     cluster = new MiniDFSCluster.Builder(Conf).NumDataNodes(DatanodeCount).Build();
     cluster.WaitActive();
     namesystem = cluster.GetNamesystem();
     bm         = namesystem.GetBlockManager();
     fs         = cluster.GetFileSystem();
 }
Example #22
0
 public virtual void SetUp()
 {
     conf    = new Configuration();
     cluster = new MiniDFSCluster.Builder(conf).NumDataNodes(Replication).Build();
     cluster.WaitActive();
     fsn  = cluster.GetNamesystem();
     hdfs = cluster.GetFileSystem();
 }
 /// <exception cref="System.IO.IOException"/>
 public virtual void StartUpCluster(long splitThreshold)
 {
     conf = new HdfsConfiguration();
     conf.SetLong(DFSConfigKeys.DfsBlockreportSplitThresholdKey, splitThreshold);
     cluster = new MiniDFSCluster.Builder(conf).NumDataNodes(ReplFactor).Build();
     fs      = cluster.GetFileSystem();
     bpid    = cluster.GetNamesystem().GetBlockPoolId();
 }
Example #24
0
 public virtual void StartUpCluster()
 {
     conf    = new HdfsConfiguration();
     cluster = new MiniDFSCluster.Builder(conf).NumDataNodes(NumDatanodes).Build();
     fs      = cluster.GetFileSystem();
     client  = fs.GetClient();
     bpid    = cluster.GetNamesystem().GetBlockPoolId();
 }
Example #25
0
 public virtual void StartUpCluster()
 {
     conf    = new HdfsConfiguration();
     cluster = new MiniDFSCluster.Builder(conf).NumDataNodes(ReplFactor).StorageTypes(
         new StorageType[] { storageType, storageType }).Build();
     fs   = cluster.GetFileSystem();
     bpid = cluster.GetNamesystem().GetBlockPoolId();
 }
Example #26
0
        public virtual void TestRetryAddBlockWhileInChooseTarget()
        {
            string            src = "/testRetryAddBlockWhileInChooseTarget";
            FSNamesystem      ns  = cluster.GetNamesystem();
            NamenodeProtocols nn  = cluster.GetNameNodeRpc();

            // create file
            nn.Create(src, FsPermission.GetFileDefault(), "clientName", new EnumSetWritable <CreateFlag
                                                                                             >(EnumSet.Of(CreateFlag.Create)), true, (short)3, 1024, null);
            // start first addBlock()
            Log.Info("Starting first addBlock for " + src);
            LocatedBlock[]        onRetryBlock = new LocatedBlock[1];
            DatanodeStorageInfo[] targets      = ns.GetNewBlockTargets(src, INodeId.GrandfatherInodeId
                                                                       , "clientName", null, null, null, onRetryBlock);
            NUnit.Framework.Assert.IsNotNull("Targets must be generated", targets);
            // run second addBlock()
            Log.Info("Starting second addBlock for " + src);
            nn.AddBlock(src, "clientName", null, null, INodeId.GrandfatherInodeId, null);
            NUnit.Framework.Assert.IsTrue("Penultimate block must be complete", CheckFileProgress
                                              (src, false));
            LocatedBlocks lbs = nn.GetBlockLocations(src, 0, long.MaxValue);

            NUnit.Framework.Assert.AreEqual("Must be one block", 1, lbs.GetLocatedBlocks().Count
                                            );
            LocatedBlock lb2 = lbs.Get(0);

            NUnit.Framework.Assert.AreEqual("Wrong replication", Replication, lb2.GetLocations
                                                ().Length);
            // continue first addBlock()
            LocatedBlock newBlock = ns.StoreAllocatedBlock(src, INodeId.GrandfatherInodeId, "clientName"
                                                           , null, targets);

            NUnit.Framework.Assert.AreEqual("Blocks are not equal", lb2.GetBlock(), newBlock.
                                            GetBlock());
            // check locations
            lbs = nn.GetBlockLocations(src, 0, long.MaxValue);
            NUnit.Framework.Assert.AreEqual("Must be one block", 1, lbs.GetLocatedBlocks().Count
                                            );
            LocatedBlock lb1 = lbs.Get(0);

            NUnit.Framework.Assert.AreEqual("Wrong replication", Replication, lb1.GetLocations
                                                ().Length);
            NUnit.Framework.Assert.AreEqual("Blocks are not equal", lb1.GetBlock(), lb2.GetBlock
                                                ());
        }
Example #27
0
        private IList <IList <int> > GetNumBlocksReport(int namesystemIdx)
        {
            IList <IList <int> > results = new AList <IList <int> >();
            string bpid = cluster.GetNamesystem(namesystemIdx).GetBlockPoolId();
            IList <IDictionary <DatanodeStorage, BlockListAsLongs> > blockReports = cluster.GetAllBlockReports
                                                                                        (bpid);

            foreach (IDictionary <DatanodeStorage, BlockListAsLongs> datanodeReport in blockReports)
            {
                IList <int> numBlocksPerDN = new AList <int>();
                foreach (BlockListAsLongs blocks in datanodeReport.Values)
                {
                    numBlocksPerDN.AddItem(blocks.GetNumberOfBlocks());
                }
                results.AddItem(numBlocksPerDN);
            }
            return(results);
        }
Example #28
0
 public virtual void SetUp()
 {
     conf = new Configuration();
     conf.SetLong(DFSConfigKeys.DfsBlockSizeKey, Blocksize);
     cluster = new MiniDFSCluster.Builder(conf).NumDataNodes(Replication).Build();
     cluster.WaitActive();
     fsdir = cluster.GetNamesystem().GetFSDirectory();
     dfs   = cluster.GetFileSystem();
 }
Example #29
0
 public virtual void SetUp()
 {
     conf    = new Configuration();
     cluster = new MiniDFSCluster.Builder(conf).NumDataNodes(Numdatanode).Build();
     cluster.WaitActive();
     fsn   = cluster.GetNamesystem();
     hdfs  = cluster.GetFileSystem();
     fsdir = fsn.GetFSDirectory();
 }
Example #30
0
        public virtual void TestDeadDatanode()
        {
            Configuration conf = new HdfsConfiguration();

            conf.SetInt(DFSConfigKeys.DfsNamenodeHeartbeatRecheckIntervalKey, 500);
            conf.SetLong(DFSConfigKeys.DfsHeartbeatIntervalKey, 1L);
            cluster = new MiniDFSCluster.Builder(conf).Build();
            cluster.WaitActive();
            string poolId = cluster.GetNamesystem().GetBlockPoolId();
            // wait for datanode to be marked live
            DataNode             dn  = cluster.GetDataNodes()[0];
            DatanodeRegistration reg = DataNodeTestUtils.GetDNRegistrationForBP(cluster.GetDataNodes
                                                                                    ()[0], poolId);

            DFSTestUtil.WaitForDatanodeState(cluster, reg.GetDatanodeUuid(), true, 20000);
            // Shutdown and wait for datanode to be marked dead
            dn.Shutdown();
            DFSTestUtil.WaitForDatanodeState(cluster, reg.GetDatanodeUuid(), false, 20000);
            DatanodeProtocol dnp = cluster.GetNameNodeRpc();

            ReceivedDeletedBlockInfo[] blocks = new ReceivedDeletedBlockInfo[] { new ReceivedDeletedBlockInfo
                                                                                     (new Block(0), ReceivedDeletedBlockInfo.BlockStatus.ReceivedBlock, null) };
            StorageReceivedDeletedBlocks[] storageBlocks = new StorageReceivedDeletedBlocks[]
            { new StorageReceivedDeletedBlocks(reg.GetDatanodeUuid(), blocks) };
            // Ensure blockReceived call from dead datanode is rejected with IOException
            try
            {
                dnp.BlockReceivedAndDeleted(reg, poolId, storageBlocks);
                NUnit.Framework.Assert.Fail("Expected IOException is not thrown");
            }
            catch (IOException)
            {
            }
            // Expected
            // Ensure blockReport from dead datanode is rejected with IOException
            StorageBlockReport[] report = new StorageBlockReport[] { new StorageBlockReport(new
                                                                                            DatanodeStorage(reg.GetDatanodeUuid()), BlockListAsLongs.Empty) };
            try
            {
                dnp.BlockReport(reg, poolId, report, new BlockReportContext(1, 0, Runtime.NanoTime
                                                                                ()));
                NUnit.Framework.Assert.Fail("Expected IOException is not thrown");
            }
            catch (IOException)
            {
            }
            // Expected
            // Ensure heartbeat from dead datanode is rejected with a command
            // that asks datanode to register again
            StorageReport[] rep = new StorageReport[] { new StorageReport(new DatanodeStorage
                                                                              (reg.GetDatanodeUuid()), false, 0, 0, 0, 0) };
            DatanodeCommand[] cmd = dnp.SendHeartbeat(reg, rep, 0L, 0L, 0, 0, 0, null).GetCommands
                                        ();
            NUnit.Framework.Assert.AreEqual(1, cmd.Length);
            NUnit.Framework.Assert.AreEqual(cmd[0].GetAction(), RegisterCommand.Register.GetAction
                                                ());
        }