/** Sets up configuration based on params */ private static boolean setup(Hashtable<String, String> curConf, Configuration argConf) { if (argConf.get("file") == null) { logger.fatal("Missing file parameter"); System.exit(1); } if (argConf.get("hdfs_base_path") == null) { logger.fatal("Missing HDFS base path, check gestore-conf.xml"); System.exit(1); } if (argConf.get("hdfs_temp_path") == null) { logger.fatal("Missing HDFS temp path, check gestore-conf.xml"); System.exit(1); } if (argConf.get("local_temp_path") == null) { logger.fatal("Missing local temp path, check gestore-conf.xml"); System.exit(1); } // Input paramaters curConf.put("run_id", argConf.get("run", "")); curConf.put("task_id", argConf.get("task", "")); curConf.put("file_id", argConf.get("file")); curConf.put("local_path", argConf.get("path", "")); curConf.put("type", argConf.get("type", "l2r")); curConf.put("timestamp_start", argConf.get("timestamp_start", "1")); curConf.put( "timestamp_stop", argConf.get("timestamp_stop", Integer.toString(Integer.MAX_VALUE))); curConf.put("delimiter", argConf.get("regex", "ID=.*")); curConf.put("taxon", argConf.get("taxon", "all")); curConf.put("intermediate", argConf.get("full_run", "false")); curConf.put("quick_add", argConf.get("quick_add", "false")); Boolean full_run = curConf.get("intermediate").matches("(?i).*true.*"); curConf.put("format", argConf.get("format", "unknown")); curConf.put("split", argConf.get("split", "1")); curConf.put("copy", argConf.get("copy", "true")); // Constants curConf.put("base_path", argConf.get("hdfs_base_path")); curConf.put("temp_path", argConf.get("hdfs_temp_path")); curConf.put("local_temp_path", argConf.get("local_temp_path")); curConf.put("db_name_files", argConf.get("hbase_file_table")); curConf.put("db_name_runs", argConf.get("hbase_run_table")); curConf.put("db_name_updates", argConf.get("hbase_db_update_table")); // Timestamps Date currentTime = new Date(); Date endDate = new Date(new Long(curConf.get("timestamp_stop"))); curConf.put("timestamp_real", Long.toString(currentTime.getTime())); return true; }
@BeforeClass public static void setUpOnce() throws Exception { miniCluster = System.getProperty("cluster.type").equals("mini"); securedCluster = System.getProperty("cluster.secured").equals("true"); System.out.println("realCluster - " + !miniCluster); System.out.println("securedCluster - " + securedCluster); Util.setLoggingThreshold("ERROR"); if (miniCluster) { if (hbase == null) { hbase = new HBaseTestingUtility(); conf = hbase.getConfiguration(); conf.set("zookeeper.session.timeout", "3600000"); conf.set("dfs.client.socket-timeout", "3600000"); if (securedCluster) { hbase.startMiniCluster(RS_COUNT); hbase.waitTableEnabled(AccessControlLists.ACL_TABLE_NAME, 30000L); admin = new HBaseAdminWrapper(conf); } else { hbase.startMiniCluster(RS_COUNT); admin = hbase.getHBaseAdmin(); } } } else { if (admin == null) { final String argsFileName = securedCluster ? "../../testClusterRealSecured.args" : "../../testClusterRealNonSecured.args"; if (!Util.isFile(argsFileName)) { throw new IllegalStateException( "You have to define args file " + argsFileName + " for tests."); } String[] testArgs = {argsFileName}; Args args = new TestArgs(testArgs); admin = HBaseClient.getAdmin(args); conf = admin.getConfiguration(); RS_COUNT = getServerNameList().size(); } } previousBalancerRunning = admin.setBalancerRunning(false, true); hConnection = HConnectionManager.createConnection(conf); USER_RW = User.createUserForTesting(conf, "rwuser", new String[0]); }
/** * Assert that getSplitEditFilesSorted returns files in expected order and that it skips * moved-aside files. * * @throws IOException */ @Test public void testGetSplitEditFilesSorted() throws IOException { FileSystem fs = FileSystem.get(util.getConfiguration()); Path regiondir = util.getDataTestDir("regiondir"); fs.delete(regiondir, true); fs.mkdirs(regiondir); Path recoverededits = WALSplitter.getRegionDirRecoveredEditsDir(regiondir); String first = WALSplitter.formatRecoveredEditsFileName(-1); createFile(fs, recoverededits, first); createFile(fs, recoverededits, WALSplitter.formatRecoveredEditsFileName(0)); createFile(fs, recoverededits, WALSplitter.formatRecoveredEditsFileName(1)); createFile(fs, recoverededits, WALSplitter.formatRecoveredEditsFileName(11)); createFile(fs, recoverededits, WALSplitter.formatRecoveredEditsFileName(2)); createFile(fs, recoverededits, WALSplitter.formatRecoveredEditsFileName(50)); String last = WALSplitter.formatRecoveredEditsFileName(Long.MAX_VALUE); createFile(fs, recoverededits, last); createFile( fs, recoverededits, Long.toString(Long.MAX_VALUE) + "." + System.currentTimeMillis()); final Configuration walConf = new Configuration(util.getConfiguration()); FSUtils.setRootDir(walConf, regiondir); (new WALFactory(walConf, null, "dummyLogName")).getWAL(new byte[] {}, null); NavigableSet<Path> files = WALSplitter.getSplitEditFilesSorted(fs, regiondir); assertEquals(7, files.size()); assertEquals(files.pollFirst().getName(), first); assertEquals(files.pollLast().getName(), last); assertEquals(files.pollFirst().getName(), WALSplitter.formatRecoveredEditsFileName(0)); assertEquals(files.pollFirst().getName(), WALSplitter.formatRecoveredEditsFileName(1)); assertEquals(files.pollFirst().getName(), WALSplitter.formatRecoveredEditsFileName(2)); assertEquals(files.pollFirst().getName(), WALSplitter.formatRecoveredEditsFileName(11)); }
/** * Assert that getSplitEditFilesSorted returns files in expected order and that it skips * moved-aside files. * * @throws IOException */ @Test public void testGetSplitEditFilesSorted() throws IOException { FileSystem fs = FileSystem.get(util.getConfiguration()); Path regiondir = util.getDataTestDir("regiondir"); fs.delete(regiondir, true); fs.mkdirs(regiondir); Path recoverededits = HLogUtil.getRegionDirRecoveredEditsDir(regiondir); String first = HLogSplitter.formatRecoveredEditsFileName(-1); createFile(fs, recoverededits, first); createFile(fs, recoverededits, HLogSplitter.formatRecoveredEditsFileName(0)); createFile(fs, recoverededits, HLogSplitter.formatRecoveredEditsFileName(1)); createFile(fs, recoverededits, HLogSplitter.formatRecoveredEditsFileName(11)); createFile(fs, recoverededits, HLogSplitter.formatRecoveredEditsFileName(2)); createFile(fs, recoverededits, HLogSplitter.formatRecoveredEditsFileName(50)); String last = HLogSplitter.formatRecoveredEditsFileName(Long.MAX_VALUE); createFile(fs, recoverededits, last); createFile( fs, recoverededits, Long.toString(Long.MAX_VALUE) + "." + System.currentTimeMillis()); HLogFactory.createHLog(fs, regiondir, "dummyLogName", util.getConfiguration()); NavigableSet<Path> files = HLogUtil.getSplitEditFilesSorted(fs, regiondir); assertEquals(7, files.size()); assertEquals(files.pollFirst().getName(), first); assertEquals(files.pollLast().getName(), last); assertEquals(files.pollFirst().getName(), HLogSplitter.formatRecoveredEditsFileName(0)); assertEquals(files.pollFirst().getName(), HLogSplitter.formatRecoveredEditsFileName(1)); assertEquals(files.pollFirst().getName(), HLogSplitter.formatRecoveredEditsFileName(2)); assertEquals(files.pollFirst().getName(), HLogSplitter.formatRecoveredEditsFileName(11)); }
private static byte[] copyKey( byte[] targetKey, int targetLength, byte[] sourceKey, int offset, int length) { if (targetLength > targetKey.length) { targetKey = new byte[targetLength]; } System.arraycopy(sourceKey, offset, targetKey, 0, length); return targetKey; }
/** * Test that we get notification if .META. moves. * * @throws IOException * @throws InterruptedException * @throws KeeperException */ @Test public void testThatIfMETAMovesWeAreNotified() throws IOException, InterruptedException, KeeperException { HConnection connection = Mockito.mock(HConnection.class); constructAndStartCatalogTracker(connection); RootLocationEditor.setRootLocation( this.watcher, new ServerName("example.com", 1234, System.currentTimeMillis())); }
@Test public void testTableInfoFileStatusComparator() { FileStatus bare = new FileStatus(0, false, 0, 0, -1, new Path("/tmp", FSTableDescriptors.TABLEINFO_NAME)); FileStatus future = new FileStatus(0, false, 0, 0, -1, new Path("/tmp/tablinfo." + System.currentTimeMillis())); FileStatus farFuture = new FileStatus( 0, false, 0, 0, -1, new Path("/tmp/tablinfo." + System.currentTimeMillis() + 1000)); FileStatus[] alist = {bare, future, farFuture}; FileStatus[] blist = {bare, farFuture, future}; FileStatus[] clist = {farFuture, bare, future}; FSTableDescriptors.FileStatusFileNameComparator c = new FSTableDescriptors.FileStatusFileNameComparator(); Arrays.sort(alist, c); Arrays.sort(blist, c); Arrays.sort(clist, c); // Now assert all sorted same in way we want. for (int i = 0; i < alist.length; i++) { assertTrue(alist[i].equals(blist[i])); assertTrue(blist[i].equals(clist[i])); assertTrue(clist[i].equals(i == 0 ? farFuture : i == 1 ? future : bare)); } }
/** * Test get of root region fails properly if nothing to connect to. * * @throws IOException * @throws InterruptedException * @throws KeeperException */ @Test public void testVerifyRootRegionLocationFails() throws IOException, InterruptedException, KeeperException { HConnection connection = Mockito.mock(HConnection.class); ConnectException connectException = new ConnectException("Connection refused"); final HRegionInterface implementation = Mockito.mock(HRegionInterface.class); Mockito.when(implementation.getRegionInfo((byte[]) Mockito.any())).thenThrow(connectException); Mockito.when( connection.getHRegionConnection( Mockito.anyString(), Mockito.anyInt(), Mockito.anyBoolean())) .thenReturn(implementation); final CatalogTracker ct = constructAndStartCatalogTracker(connection); RootLocationEditor.setRootLocation( this.watcher, new ServerName("example.com", 1234, System.currentTimeMillis())); Assert.assertFalse(ct.verifyRootRegionLocation(100)); }
public HostRankHBaseTest() { super(HostRankHBaseTest.class.getName()); // Let's set up the hbase root directory. Configuration conf = HBaseConfiguration.create(); try { FileSystem fs = FileSystem.get(conf); String randomStr = UUID.randomUUID().toString(); String tmpdir = System.getProperty("java.io.tmpdir") + "/" + randomStr + "/"; hbaseRootdir = fs.makeQualified(new Path(tmpdir)); conf.set(HConstants.HBASE_DIR, hbaseRootdir.toString()); fs.mkdirs(hbaseRootdir); } catch (IOException ioe) { fail("Could not create hbase root directory."); } // Start the test utility. testUtil = new HBaseTestingUtility(conf); }
@Test public void testHostRank() throws Exception { if (System.getProperty("prop.mapred.job.tracker") != null) { if (LOG.isInfoEnabled()) LOG.info("testHBaseInputOutput: Ignore this test if not local mode."); return; } File jarTest = new File(System.getProperty("prop.jarLocation")); if (!jarTest.exists()) { fail( "Could not find Giraph jar at " + "location specified by 'prop.jarLocation'. " + "Make sure you built the main Giraph artifact?."); } MiniHBaseCluster cluster = null; MiniZooKeeperCluster zkCluster = null; FileSystem fs = null; try { // using the restart method allows us to avoid having the hbase // root directory overwritten by /home/$username zkCluster = testUtil.startMiniZKCluster(); testUtil.restartHBaseCluster(2); cluster = testUtil.getMiniHBaseCluster(); final byte[] OL_BYTES = Bytes.toBytes("ol"); final byte[] S_BYTES = Bytes.toBytes("s"); final byte[] METADATA_BYTES = Bytes.toBytes("mtdt"); final byte[] HR_BYTES = Bytes.toBytes("_hr_"); final byte[] TAB = Bytes.toBytes(TABLE_NAME); Configuration conf = cluster.getConfiguration(); HTableDescriptor desc = new HTableDescriptor(TAB); desc.addFamily(new HColumnDescriptor(OL_BYTES)); desc.addFamily(new HColumnDescriptor(S_BYTES)); desc.addFamily(new HColumnDescriptor(METADATA_BYTES)); HBaseAdmin hbaseAdmin = new HBaseAdmin(conf); if (hbaseAdmin.isTableAvailable(TABLE_NAME)) { hbaseAdmin.disableTable(TABLE_NAME); hbaseAdmin.deleteTable(TABLE_NAME); } hbaseAdmin.createTable(desc); /** * Enter the initial data (a,b), (b,c), (a,c) a = 1.0 - google b = 1.0 - yahoo c = 1.0 - bing */ HTable table = new HTable(conf, TABLE_NAME); Put p1 = new Put(Bytes.toBytes("com.google.www")); p1.add(OL_BYTES, Bytes.toBytes("www.yahoo.com"), Bytes.toBytes("ab")); Put p2 = new Put(Bytes.toBytes("com.google.www")); p2.add(OL_BYTES, Bytes.toBytes("www.bing.com"), Bytes.toBytes("ac")); p2.add(OL_BYTES, Bytes.toBytes("www.bing.com"), Bytes.toBytes("invalid1")); p2.add(OL_BYTES, Bytes.toBytes("www.google.com"), Bytes.toBytes("invalid2")); Put p3 = new Put(Bytes.toBytes("com.yahoo.www")); p3.add(OL_BYTES, Bytes.toBytes("www.bing.com"), Bytes.toBytes("bc")); // p3.add(OL_BYTES, Bytes.toBytes(""), Bytes.toBytes("invalid4")); Put p4 = new Put(Bytes.toBytes("com.bing.www")); // TODO: Handle below case. use apache isValid method. p4.add(OL_BYTES, Bytes.toBytes("http://invalidurl"), Bytes.toBytes("invalid5")); p4.add(S_BYTES, S_BYTES, Bytes.toBytes(10.0d)); Put p5 = new Put(Bytes.toBytes("dummy")); p5.add(S_BYTES, S_BYTES, Bytes.toBytes(10.0d)); table.put(p1); table.put(p2); table.put(p3); table.put(p4); table.put(p5); // Set Giraph configuration // now operate over HBase using Vertex I/O formats conf.set(TableInputFormat.INPUT_TABLE, TABLE_NAME); conf.set(TableOutputFormat.OUTPUT_TABLE, TABLE_NAME); // Start the giraph job GiraphJob giraphJob = new GiraphJob(conf, BspCase.getCallingMethodName()); GiraphConfiguration giraphConf = giraphJob.getConfiguration(); giraphConf.setZooKeeperConfiguration(cluster.getMaster().getZooKeeper().getQuorum()); setupConfiguration(giraphJob); giraphConf.setComputationClass(LinkRankComputation.class); giraphConf.setMasterComputeClass(LinkRankVertexMasterCompute.class); giraphConf.setOutEdgesClass(ByteArrayEdges.class); giraphConf.setVertexInputFormatClass(Nutch2HostInputFormat.class); giraphConf.setVertexOutputFormatClass(Nutch2HostOutputFormat.class); giraphConf.setInt("giraph.linkRank.superstepCount", 10); giraphConf.setInt("giraph.linkRank.scale", 10); giraphConf.set("giraph.linkRank.family", "mtdt"); giraphConf.set("giraph.linkRank.qualifier", "_hr_"); giraphConf.setVertexInputFilterClass(HostRankVertexFilter.class); assertTrue(giraphJob.run(true)); if (LOG.isInfoEnabled()) LOG.info("Giraph job successful. Checking output qualifier."); /** Check the results * */ Result result; String key; byte[] calculatedScoreByte; HashMap expectedValues = new HashMap<String, Double>(); expectedValues.put("com.google.www", 1.3515060339386287d); expectedValues.put("com.yahoo.www", 4.144902009567587d); expectedValues.put("com.bing.www", 9.063893290511482d); for (Object keyObject : expectedValues.keySet()) { key = keyObject.toString(); result = table.get(new Get(key.getBytes())); calculatedScoreByte = result.getValue(METADATA_BYTES, HR_BYTES); assertNotNull(calculatedScoreByte); assertTrue(calculatedScoreByte.length > 0); Assert.assertEquals( "Scores are not the same", (Double) expectedValues.get(key), Bytes.toDouble(calculatedScoreByte), DELTA); } } finally { if (cluster != null) { cluster.shutdown(); } if (zkCluster != null) { zkCluster.shutdown(); } // clean test files if (fs != null) { fs.delete(hbaseRootdir); } } }
@Test public void testLogCleaning() throws Exception { Configuration conf = TEST_UTIL.getConfiguration(); // set TTL long ttl = 10000; conf.setLong("hbase.master.logcleaner.ttl", ttl); conf.setBoolean(HConstants.REPLICATION_ENABLE_KEY, HConstants.REPLICATION_ENABLE_DEFAULT); Replication.decorateMasterConfiguration(conf); Server server = new DummyServer(); ReplicationQueues repQueues = ReplicationFactory.getReplicationQueues(server.getZooKeeper(), conf, server); repQueues.init(server.getServerName().toString()); final Path oldLogDir = new Path(TEST_UTIL.getDataTestDir(), HConstants.HREGION_OLDLOGDIR_NAME); String fakeMachineName = URLEncoder.encode(server.getServerName().toString(), "UTF8"); final FileSystem fs = FileSystem.get(conf); // Create 2 invalid files, 1 "recent" file, 1 very new file and 30 old files long now = System.currentTimeMillis(); fs.delete(oldLogDir, true); fs.mkdirs(oldLogDir); // Case 1: 2 invalid files, which would be deleted directly fs.createNewFile(new Path(oldLogDir, "a")); fs.createNewFile(new Path(oldLogDir, fakeMachineName + "." + "a")); // Case 2: 1 "recent" file, not even deletable for the first log cleaner // (TimeToLiveLogCleaner), so we are not going down the chain System.out.println("Now is: " + now); for (int i = 1; i < 31; i++) { // Case 3: old files which would be deletable for the first log cleaner // (TimeToLiveLogCleaner), and also for the second (ReplicationLogCleaner) Path fileName = new Path(oldLogDir, fakeMachineName + "." + (now - i)); fs.createNewFile(fileName); // Case 4: put 3 old log files in ZK indicating that they are scheduled // for replication so these files would pass the first log cleaner // (TimeToLiveLogCleaner) but would be rejected by the second // (ReplicationLogCleaner) if (i % (30 / 3) == 1) { repQueues.addLog(fakeMachineName, fileName.getName()); System.out.println("Replication log file: " + fileName); } } // sleep for sometime to get newer modifcation time Thread.sleep(ttl); fs.createNewFile(new Path(oldLogDir, fakeMachineName + "." + now)); // Case 2: 1 newer file, not even deletable for the first log cleaner // (TimeToLiveLogCleaner), so we are not going down the chain fs.createNewFile(new Path(oldLogDir, fakeMachineName + "." + (now + 10000))); for (FileStatus stat : fs.listStatus(oldLogDir)) { System.out.println(stat.getPath().toString()); } assertEquals(34, fs.listStatus(oldLogDir).length); LogCleaner cleaner = new LogCleaner(1000, server, conf, fs, oldLogDir); cleaner.chore(); // We end up with the current log file, a newer one and the 3 old log // files which are scheduled for replication TEST_UTIL.waitFor( 1000, new Waiter.Predicate<Exception>() { @Override public boolean evaluate() throws Exception { return 5 == fs.listStatus(oldLogDir).length; } }); for (FileStatus file : fs.listStatus(oldLogDir)) { System.out.println("Kept log files: " + file.getPath().getName()); } }
public int run(String[] args) throws Exception { // printUsage(); /* * SETUP */ Configuration argConf = getConf(); Hashtable<String, String> confArg = new Hashtable<String, String>(); setup(confArg, argConf); Date currentTime = new Date(); Date endDate = new Date(new Long(confArg.get("timestamp_stop"))); Boolean full_run = confArg.get("intermediate").matches("(?i).*true.*"); Boolean quick_add = confArg.get("quick_add").matches("(?i).*true.*"); logger.info("Running GeStore"); // ZooKeeper setup Configuration config = HBaseConfiguration.create(); zkWatcher = new ZooKeeperWatcher(config, "Testing", new HBaseAdmin(config)); zkInstance = new ZooKeeper( ZKConfig.getZKQuorumServersString(config), config.getInt("zookeeper.session.timeout", -1), zkWatcher); if (!confArg.get("task_id").isEmpty()) { confArg.put("temp_path", confArg.get("temp_path") + confArg.get("task_id")); } String lockRequest = confArg.get("file_id"); if (!confArg.get("run_id").isEmpty()) lockRequest = lockRequest + "_" + confArg.get("run_id") + "_"; if (!confArg.get("task_id").isEmpty()) lockRequest = lockRequest + "_" + confArg.get("task_id") + "_"; // Get type of movement toFrom type_move = checkArgs(confArg); if (type_move == toFrom.LOCAL2REMOTE && !confArg.get("format").equals("unknown")) { List<String> arguments = new ArrayList<String>(); arguments.add("-Dinput=" + confArg.get("local_path")); arguments.add("-Dtable=" + confArg.get("file_id")); arguments.add("-Dtimestamp=" + confArg.get("timestamp_stop")); arguments.add("-Dtype=" + confArg.get("format")); arguments.add("-Dtarget_dir=" + confArg.get("base_path") + "_" + confArg.get("file_id")); arguments.add("-Dtemp_hdfs_path=" + confArg.get("temp_path")); arguments.add("-Drun_id=" + confArg.get("run_id")); if (!confArg.get("run_id").isEmpty()) arguments.add("-Drun_id=" + confArg.get("run_id")); if (!confArg.get("task_id").isEmpty()) arguments.add("-Dtask_id=" + confArg.get("task_id")); if (quick_add) arguments.add("-Dquick_add=" + confArg.get("quick_add")); String lockName = lock(lockRequest); String[] argumentString = arguments.toArray(new String[arguments.size()]); adddb.main(argumentString); unlock(lockName); System.exit(0); } // Database registration dbutil db_util = new dbutil(config); db_util.register_database(confArg.get("db_name_files"), true); db_util.register_database(confArg.get("db_name_runs"), true); db_util.register_database(confArg.get("db_name_updates"), true); FileSystem hdfs = FileSystem.get(config); FileSystem localFS = FileSystem.getLocal(config); // Get source type confArg.put("source", getSource(db_util, confArg.get("db_name_files"), confArg.get("file_id"))); confArg.put( "database", isDatabase(db_util, confArg.get("db_name_files"), confArg.get("file_id"))); if (!confArg.get("source").equals("local") && type_move == toFrom.REMOTE2LOCAL && !confArg.get("timestamp_stop").equals(Integer.toString(Integer.MAX_VALUE))) { confArg.put("timestamp_stop", Long.toString(latestVersion(confArg, db_util))); } /* * Get previous timestamp */ Get run_id_get = new Get(confArg.get("run_id").getBytes()); Result run_get = db_util.doGet(confArg.get("db_name_runs"), run_id_get); KeyValue run_file_prev = run_get.getColumnLatest( "d".getBytes(), (confArg.get("file_id") + "_db_timestamp").getBytes()); String last_timestamp = new String("0"); if (null != run_file_prev && !confArg.get("source").equals("local")) { long last_timestamp_real = run_file_prev.getTimestamp(); Long current_timestamp = new Long(confArg.get("timestamp_real")); if ((current_timestamp - last_timestamp_real) > 36000) { last_timestamp = new String(run_file_prev.getValue()); Integer lastTimestamp = new Integer(last_timestamp); lastTimestamp += 1; last_timestamp = lastTimestamp.toString(); logger.info("Last timestamp: " + last_timestamp + " End data: " + endDate); Date last_run = new Date(run_file_prev.getTimestamp()); if (last_run.before(endDate) && !full_run) { confArg.put("timestamp_start", last_timestamp); } } } Integer tse = new Integer(confArg.get("timestamp_stop")); Integer tss = new Integer(confArg.get("timestamp_start")); if (tss > tse) { logger.info("No new version of requested file."); return 0; } /* * Generate file */ String lockName = lock(lockRequest); Get file_id_get = new Get(confArg.get("file_id").getBytes()); Result file_get = db_util.doGet(confArg.get("db_name_files"), file_id_get); if (!file_get.isEmpty()) { boolean found = hasFile( db_util, hdfs, confArg.get("db_name_files"), confArg.get("file_id"), getFullPath(confArg)); if (confArg.get("source").equals("fullfile")) { found = false; } String filenames_put = getFileNames( db_util, confArg.get("db_name_files"), confArg.get("file_id"), getFullPath(confArg)); // Filename not found in file database if (!found && type_move == toFrom.REMOTE2LOCAL) { if (!confArg.get("source").equals("local")) { // Generate intermediate file if (getFile(hdfs, confArg, db_util) == null) { unlock(lockName); return 1; } // Put generated file into file database if (!confArg.get("format").equals("fullfile")) { putFileEntry( db_util, hdfs, confArg.get("db_name_files"), confArg.get("file_id"), confArg.get("full_file_name"), confArg.get("source")); } } else { logger.warn("Remote file not found, and cannot be generated! File: " + confArg); unlock(lockName); return 1; } } } else { if (type_move == toFrom.REMOTE2LOCAL) { logger.warn("Remote file not found, and cannot be generated."); unlock(lockName); return 1; } } /* * Copy file * Update tables */ if (type_move == toFrom.LOCAL2REMOTE) { if (!confArg.get("format").equals("fullfile")) { putFileEntry( db_util, hdfs, confArg.get("db_name_files"), confArg.get("file_id"), getFullPath(confArg), confArg.get("source")); } putRunEntry( db_util, confArg.get("db_name_runs"), confArg.get("run_id"), confArg.get("file_id"), confArg.get("type"), confArg.get("timestamp_real"), confArg.get("timestamp_stop"), getFullPath(confArg), confArg.get("delimiter")); hdfs.copyFromLocalFile(new Path(confArg.get("local_path")), new Path(getFullPath(confArg))); } else if (type_move == toFrom.REMOTE2LOCAL) { FileStatus[] files = hdfs.globStatus(new Path(getFullPath(confArg) + "*")); putRunEntry( db_util, confArg.get("db_name_runs"), confArg.get("run_id"), confArg.get("file_id"), confArg.get("type"), confArg.get("timestamp_real"), confArg.get("timestamp_stop"), getFullPath(confArg), confArg.get("delimiter")); unlock(lockName); for (FileStatus file : files) { Path cur_file = file.getPath(); Path cur_local_path = new Path(new String(confArg.get("local_path") + confArg.get("file_id"))); String suffix = getSuffix(getFileName(confArg), cur_file.getName()); if (suffix.length() > 0) { cur_local_path = cur_local_path.suffix(new String("." + suffix)); } if (confArg.get("copy").equals("true")) { String crc = hdfs.getFileChecksum(cur_file).toString(); if (checksumLocalTest(cur_local_path, crc)) { continue; } else { hdfs.copyToLocalFile(cur_file, cur_local_path); writeChecksum(cur_local_path, crc); } } else { System.out.println(cur_local_path + "\t" + cur_file); } } } unlock(lockName); return 0; }
public static void main(String[] args) throws Exception { int result = ToolRunner.run(new Configuration(), new move(), args); System.exit(result); }
/** * Test that MetaReader will ride over server throwing "Server not running" IOEs. * * @see https://issues.apache.org/jira/browse/HBASE-3446 * @throws IOException * @throws InterruptedException */ @Test public void testRideOverServerNotRunning() throws IOException, InterruptedException { // Need a zk watcher. ZooKeeperWatcher zkw = new ZooKeeperWatcher( UTIL.getConfiguration(), this.getClass().getSimpleName(), ABORTABLE, true); // This is a servername we use in a few places below. ServerName sn = new ServerName("example.com", 1234, System.currentTimeMillis()); HConnection connection = null; CatalogTracker ct = null; try { // Mock an HRegionInterface. Our mock implementation will fail a few // times when we go to open a scanner. final HRegionInterface implementation = Mockito.mock(HRegionInterface.class); // When openScanner called throw IOE 'Server not running' a few times // before we return a scanner id. Whats WEIRD is that these // exceptions do not show in the log because they are caught and only // printed if we FAIL. We eventually succeed after retry so these don't // show. We will know if they happened or not because we will ask // mockito at the end of this test to verify that openscanner was indeed // called the wanted number of times. final long scannerid = 123L; Mockito.when(implementation.openScanner((byte[]) Mockito.any(), (Scan) Mockito.any())) .thenThrow(new IOException("Server not running (1 of 3)")) .thenThrow(new IOException("Server not running (2 of 3)")) .thenThrow(new IOException("Server not running (3 of 3)")) .thenReturn(scannerid); // Make it so a verifiable answer comes back when next is called. Return // the verifiable answer and then a null so we stop scanning. Our // verifiable answer is something that looks like a row in META with // a server and startcode that is that of the above defined servername. List<KeyValue> kvs = new ArrayList<KeyValue>(); final byte[] rowToVerify = Bytes.toBytes("rowToVerify"); kvs.add( new KeyValue( rowToVerify, HConstants.CATALOG_FAMILY, HConstants.REGIONINFO_QUALIFIER, Writables.getBytes(HRegionInfo.FIRST_META_REGIONINFO))); kvs.add( new KeyValue( rowToVerify, HConstants.CATALOG_FAMILY, HConstants.SERVER_QUALIFIER, Bytes.toBytes(sn.getHostAndPort()))); kvs.add( new KeyValue( rowToVerify, HConstants.CATALOG_FAMILY, HConstants.STARTCODE_QUALIFIER, Bytes.toBytes(sn.getStartcode()))); final Result[] result = new Result[] {new Result(kvs)}; Mockito.when(implementation.next(Mockito.anyLong(), Mockito.anyInt())) .thenReturn(result) .thenReturn(null); // Associate a spied-upon HConnection with UTIL.getConfiguration. Need // to shove this in here first so it gets picked up all over; e.g. by // HTable. connection = HConnectionTestingUtility.getSpiedConnection(UTIL.getConfiguration()); // Fix the location lookup so it 'works' though no network. First // make an 'any location' object. final HRegionLocation anyLocation = new HRegionLocation(HRegionInfo.FIRST_META_REGIONINFO, sn.getHostname(), sn.getPort()); // Return the any location object when locateRegion is called in HTable // constructor and when its called by ServerCallable (it uses getRegionLocation). // The ugly format below comes of 'Important gotcha on spying real objects!' from // http://mockito.googlecode.com/svn/branches/1.6/javadoc/org/mockito/Mockito.html Mockito.doReturn(anyLocation) .when(connection) .locateRegion((byte[]) Mockito.any(), (byte[]) Mockito.any()); Mockito.doReturn(anyLocation) .when(connection) .getRegionLocation((byte[]) Mockito.any(), (byte[]) Mockito.any(), Mockito.anyBoolean()); // Now shove our HRI implementation into the spied-upon connection. Mockito.doReturn(implementation) .when(connection) .getHRegionConnection(Mockito.anyString(), Mockito.anyInt()); // Now start up the catalogtracker with our doctored Connection. ct = new CatalogTracker(zkw, null, connection, ABORTABLE, 0); ct.start(); // Scan meta for user tables and verify we got back expected answer. NavigableMap<HRegionInfo, Result> hris = MetaReader.getServerUserRegions(ct, sn); assertTrue(hris.size() == 1); assertTrue(hris.firstEntry().getKey().equals(HRegionInfo.FIRST_META_REGIONINFO)); assertTrue(Bytes.equals(rowToVerify, hris.firstEntry().getValue().getRow())); // Finally verify that openscanner was called four times -- three times // with exception and then on 4th attempt we succeed. Mockito.verify(implementation, Mockito.times(4)) .openScanner((byte[]) Mockito.any(), (Scan) Mockito.any()); } finally { if (ct != null) ct.stop(); HConnectionManager.deleteConnection(UTIL.getConfiguration(), true); zkw.close(); } }
/** Test {@link CatalogTracker} */ @Category(MediumTests.class) public class TestCatalogTracker { private static final Log LOG = LogFactory.getLog(TestCatalogTracker.class); private static final HBaseTestingUtility UTIL = new HBaseTestingUtility(); private static final ServerName SN = new ServerName("example.org", 1234, System.currentTimeMillis()); private ZooKeeperWatcher watcher; private Abortable abortable; @BeforeClass public static void beforeClass() throws Exception { // Set this down so tests run quicker UTIL.getConfiguration().setInt(HConstants.HBASE_CLIENT_RETRIES_NUMBER, 3); UTIL.startMiniZKCluster(); } @AfterClass public static void afterClass() throws IOException { UTIL.getZkCluster().shutdown(); } @Before public void before() throws IOException { this.abortable = new Abortable() { @Override public void abort(String why, Throwable e) { LOG.info(why, e); } @Override public boolean isAborted() { return false; } }; this.watcher = new ZooKeeperWatcher( UTIL.getConfiguration(), this.getClass().getSimpleName(), this.abortable, true); } @After public void after() { try { // Clean out root location or later tests will be confused... they presume // start fresh in zk. RootLocationEditor.deleteRootLocation(this.watcher); } catch (KeeperException e) { LOG.warn("Unable to delete root location", e); } // Clear out our doctored connection or could mess up subsequent tests. HConnectionManager.deleteConnection(UTIL.getConfiguration()); this.watcher.close(); } private CatalogTracker constructAndStartCatalogTracker(final HConnection c) throws IOException, InterruptedException { CatalogTracker ct = new CatalogTracker(this.watcher, UTIL.getConfiguration(), c, this.abortable); ct.start(); return ct; } /** * Test that we get notification if .META. moves. * * @throws IOException * @throws InterruptedException * @throws KeeperException */ @Test public void testThatIfMETAMovesWeAreNotified() throws IOException, InterruptedException, KeeperException { HConnection connection = Mockito.mock(HConnection.class); constructAndStartCatalogTracker(connection); RootLocationEditor.setRootLocation( this.watcher, new ServerName("example.com", 1234, System.currentTimeMillis())); } /** * Test interruptable while blocking wait on root and meta. * * @throws IOException * @throws InterruptedException */ @Test public void testInterruptWaitOnMetaAndRoot() throws IOException, InterruptedException { HRegionInterface implementation = Mockito.mock(HRegionInterface.class); HConnection connection = mockConnection(implementation); final CatalogTracker ct = constructAndStartCatalogTracker(connection); ServerName hsa = ct.getRootLocation(); Assert.assertNull(hsa); ServerName meta = ct.getMetaLocation(); Assert.assertNull(meta); Thread t = new Thread() { @Override public void run() { try { ct.waitForMeta(); } catch (InterruptedException e) { throw new RuntimeException("Interrupted", e); } } }; t.start(); while (!t.isAlive()) Threads.sleep(1); Threads.sleep(1); assertTrue(t.isAlive()); ct.stop(); // Join the thread... should exit shortly. t.join(); } /** * Test for HBASE-4288. Throw an IOE when trying to verify meta region and prove it doesn't cause * master shutdown. * * @see <a href="https://issues.apache.org/jira/browse/HBASE-4288">HBASE-4288</a> * @throws IOException * @throws InterruptedException * @throws KeeperException */ @Test public void testServerNotRunningIOException() throws IOException, InterruptedException, KeeperException { // Mock an HRegionInterface. final HRegionInterface implementation = Mockito.mock(HRegionInterface.class); HConnection connection = mockConnection(implementation); // If a 'getRegionInfo' is called on mocked HRegionInterface, throw IOE // the first time. 'Succeed' the second time we are called. Mockito.when(implementation.getRegionInfo((byte[]) Mockito.any())) .thenThrow(new IOException("Server not running, aborting")) .thenReturn(new HRegionInfo()); // After we encounter the above 'Server not running', we should catch the // IOE and go into retrying for the meta mode. We'll do gets on -ROOT- to // get new meta location. Return something so this 'get' succeeds // (here we mock up getRegionServerWithRetries, the wrapper around // the actual get). // TODO: Refactor. This method has been moved out of HConnection. // It works for now but has been deprecated. Mockito.when(connection.getRegionServerWithRetries((ServerCallable<Result>) Mockito.any())) .thenReturn(getMetaTableRowResult()); // Now start up the catalogtracker with our doctored Connection. final CatalogTracker ct = constructAndStartCatalogTracker(connection); try { // Set a location for root and meta. RootLocationEditor.setRootLocation(this.watcher, SN); ct.setMetaLocation(SN); // Call the method that HBASE-4288 calls. It will try and verify the // meta location and will fail on first attempt then go into a long wait. // So, do this in a thread and then reset meta location to break it out // of its wait after a bit of time. final AtomicBoolean metaSet = new AtomicBoolean(false); final CountDownLatch latch = new CountDownLatch(1); Thread t = new Thread() { @Override public void run() { try { latch.countDown(); metaSet.set(ct.waitForMeta(100000) != null); } catch (Exception e) { throw new RuntimeException(e); } } }; t.start(); latch.await(); Threads.sleep(1); // Now reset the meta as though it were redeployed. ct.setMetaLocation(SN); t.join(); Assert.assertTrue(metaSet.get()); } finally { // Clean out root and meta locations or later tests will be confused... // they presume start fresh in zk. ct.resetMetaLocation(); } } private void testVerifyMetaRegionLocationWithException(Exception ex) throws IOException, InterruptedException, KeeperException { // Mock an HRegionInterface. final HRegionInterface implementation = Mockito.mock(HRegionInterface.class); HConnection connection = mockConnection(implementation); // If a 'get' is called on mocked interface, throw connection refused. Mockito.when(implementation.get((byte[]) Mockito.any(), (Get) Mockito.any())).thenThrow(ex); // Now start up the catalogtracker with our doctored Connection. final CatalogTracker ct = constructAndStartCatalogTracker(connection); RootLocationEditor.setRootLocation(this.watcher, SN); long timeout = UTIL.getConfiguration().getLong("hbase.catalog.verification.timeout", 1000); Assert.assertFalse(ct.verifyMetaRegionLocation(timeout)); } /** * Test we survive a connection refused {@link ConnectException} * * @throws IOException * @throws InterruptedException * @throws KeeperException */ @Test public void testGetMetaServerConnectionFails() throws IOException, InterruptedException, KeeperException { testVerifyMetaRegionLocationWithException(new ConnectException("Connection refused")); } /** * Test that verifyMetaRegionLocation properly handles getting a ServerNotRunningException. See * HBASE-4470. Note this doesn't check the exact exception thrown in the HBASE-4470 as there it is * thrown from getHConnection() and here it is thrown from get() -- but those are both called from * the same function anyway, and this way is less invasive than throwing from getHConnection would * be. * * @throws IOException * @throws InterruptedException * @throws KeeperException */ @Test public void testVerifyMetaRegionServerNotRunning() throws IOException, InterruptedException, KeeperException { testVerifyMetaRegionLocationWithException(new ServerNotRunningYetException("mock")); } /** * Test get of root region fails properly if nothing to connect to. * * @throws IOException * @throws InterruptedException * @throws KeeperException */ @Test public void testVerifyRootRegionLocationFails() throws IOException, InterruptedException, KeeperException { HConnection connection = Mockito.mock(HConnection.class); ConnectException connectException = new ConnectException("Connection refused"); final HRegionInterface implementation = Mockito.mock(HRegionInterface.class); Mockito.when(implementation.getRegionInfo((byte[]) Mockito.any())).thenThrow(connectException); Mockito.when( connection.getHRegionConnection( Mockito.anyString(), Mockito.anyInt(), Mockito.anyBoolean())) .thenReturn(implementation); final CatalogTracker ct = constructAndStartCatalogTracker(connection); RootLocationEditor.setRootLocation( this.watcher, new ServerName("example.com", 1234, System.currentTimeMillis())); Assert.assertFalse(ct.verifyRootRegionLocation(100)); } @Test(expected = NotAllMetaRegionsOnlineException.class) public void testTimeoutWaitForRoot() throws IOException, InterruptedException { HConnection connection = Mockito.mock(HConnection.class); final CatalogTracker ct = constructAndStartCatalogTracker(connection); ct.waitForRoot(100); } @Test(expected = RetriesExhaustedException.class) public void testTimeoutWaitForMeta() throws IOException, InterruptedException { HConnection connection = HConnectionTestingUtility.getMockedConnection(UTIL.getConfiguration()); final CatalogTracker ct = constructAndStartCatalogTracker(connection); ct.waitForMeta(100); } /** * Test waiting on root w/ no timeout specified. * * @throws IOException * @throws InterruptedException * @throws KeeperException */ @Test public void testNoTimeoutWaitForRoot() throws IOException, InterruptedException, KeeperException { HConnection connection = Mockito.mock(HConnection.class); final CatalogTracker ct = constructAndStartCatalogTracker(connection); ServerName hsa = ct.getRootLocation(); Assert.assertNull(hsa); // Now test waiting on root location getting set. Thread t = new WaitOnMetaThread(ct); startWaitAliveThenWaitItLives(t, 1000); // Set a root location. hsa = setRootLocation(); // Join the thread... should exit shortly. t.join(); // Now root is available. Assert.assertTrue(ct.getRootLocation().equals(hsa)); } private ServerName setRootLocation() throws KeeperException { RootLocationEditor.setRootLocation(this.watcher, SN); return SN; } /** * Test waiting on meta w/ no timeout specified. * * @throws Exception */ @Ignore // Can't make it work reliably on all platforms; mockito gets confused // Throwing: org.mockito.exceptions.misusing.WrongTypeOfReturnValue: // Result cannot be returned by locateRegion() // If you plug locateRegion, it then throws for incCounter, and if you plug // that ... and so one. @Test public void testNoTimeoutWaitForMeta() throws Exception { // Mock an HConnection and a HRegionInterface implementation. Have the // HConnection return the HRI. Have the HRI return a few mocked up responses // to make our test work. // Mock an HRegionInterface. final HRegionInterface implementation = Mockito.mock(HRegionInterface.class); HConnection connection = mockConnection(implementation); // Now the ct is up... set into the mocks some answers that make it look // like things have been getting assigned. Make it so we'll return a // location (no matter what the Get is). Same for getHRegionInfo -- always // just return the meta region. final Result result = getMetaTableRowResult(); // TODO: Refactor. This method has been moved out of HConnection. // It works for now but has been deprecated. Mockito.when(connection.getRegionServerWithRetries((ServerCallable<Result>) Mockito.any())) .thenReturn(result); Mockito.when(implementation.getRegionInfo((byte[]) Mockito.any())) .thenReturn(HRegionInfo.FIRST_META_REGIONINFO); final CatalogTracker ct = constructAndStartCatalogTracker(connection); ServerName hsa = ct.getMetaLocation(); Assert.assertNull(hsa); // Now test waiting on meta location getting set. Thread t = new WaitOnMetaThread(ct) { @Override void doWaiting() throws InterruptedException { this.ct.waitForMeta(); } }; startWaitAliveThenWaitItLives(t, 1000); // This should trigger wake up of meta wait (Its the removal of the meta // region unassigned node that triggers catalogtrackers that a meta has // been assigned). String node = ct.getMetaNodeTracker().getNode(); ZKUtil.createAndFailSilent(this.watcher, node); MetaEditor.updateMetaLocation(ct, HRegionInfo.FIRST_META_REGIONINFO, SN); ZKUtil.deleteNode(this.watcher, node); // Go get the new meta location. waitForMeta gets and verifies meta. Assert.assertTrue(ct.waitForMeta(10000).equals(SN)); // Join the thread... should exit shortly. t.join(); // Now meta is available. Assert.assertTrue(ct.waitForMeta(10000).equals(SN)); } /** * @param implementation An {@link HRegionInterface} instance; you'll likely want to pass a mocked * HRS; can be null. * @return Mock up a connection that returns a {@link org.apache.hadoop.conf.Configuration} when * {@link HConnection#getConfiguration()} is called, a 'location' when {@link * HConnection#getRegionLocation(byte[], byte[], boolean)} is called, and that returns the * passed {@link HRegionInterface} instance when {@link * HConnection#getHRegionConnection(String, int)} is called (Be sure call {@link * HConnectionManager#deleteConnection(org.apache.hadoop.conf.Configuration)} when done with * this mocked Connection. * @throws IOException */ private HConnection mockConnection(final HRegionInterface implementation) throws IOException { HConnection connection = HConnectionTestingUtility.getMockedConnection(UTIL.getConfiguration()); Mockito.doNothing().when(connection).close(); // Make it so we return any old location when asked. final HRegionLocation anyLocation = new HRegionLocation(HRegionInfo.FIRST_META_REGIONINFO, SN.getHostname(), SN.getPort()); Mockito.when( connection.getRegionLocation( (byte[]) Mockito.any(), (byte[]) Mockito.any(), Mockito.anyBoolean())) .thenReturn(anyLocation); Mockito.when(connection.locateRegion((byte[]) Mockito.any(), (byte[]) Mockito.any())) .thenReturn(anyLocation); if (implementation != null) { // If a call to getHRegionConnection, return this implementation. Mockito.when(connection.getHRegionConnection(Mockito.anyString(), Mockito.anyInt())) .thenReturn(implementation); } return connection; } /** * @return A mocked up Result that fakes a Get on a row in the <code>.META.</code> table. * @throws IOException */ private Result getMetaTableRowResult() throws IOException { List<KeyValue> kvs = new ArrayList<KeyValue>(); kvs.add( new KeyValue( HConstants.EMPTY_BYTE_ARRAY, HConstants.CATALOG_FAMILY, HConstants.REGIONINFO_QUALIFIER, Writables.getBytes(HRegionInfo.FIRST_META_REGIONINFO))); kvs.add( new KeyValue( HConstants.EMPTY_BYTE_ARRAY, HConstants.CATALOG_FAMILY, HConstants.SERVER_QUALIFIER, Bytes.toBytes(SN.getHostAndPort()))); kvs.add( new KeyValue( HConstants.EMPTY_BYTE_ARRAY, HConstants.CATALOG_FAMILY, HConstants.STARTCODE_QUALIFIER, Bytes.toBytes(SN.getStartcode()))); return new Result(kvs); } private void startWaitAliveThenWaitItLives(final Thread t, final int ms) { t.start(); while (!t.isAlive()) { // Wait } // Wait one second. Threads.sleep(ms); Assert.assertTrue("Assert " + t.getName() + " still waiting", t.isAlive()); } class CountingProgressable implements Progressable { final AtomicInteger counter = new AtomicInteger(0); @Override public void progress() { this.counter.incrementAndGet(); } } /** Wait on META. Default is wait on -ROOT-. */ class WaitOnMetaThread extends Thread { final CatalogTracker ct; WaitOnMetaThread(final CatalogTracker ct) { super("WaitOnMeta"); this.ct = ct; } @Override public void run() { try { doWaiting(); } catch (InterruptedException e) { throw new RuntimeException("Failed wait", e); } LOG.info("Exiting " + getName()); } void doWaiting() throws InterruptedException { this.ct.waitForRoot(); } } @org.junit.Rule public org.apache.hadoop.hbase.ResourceCheckerJUnitRule cu = new org.apache.hadoop.hbase.ResourceCheckerJUnitRule(); }