[120] | 1 | /** |
---|
| 2 | * Licensed to the Apache Software Foundation (ASF) under one |
---|
| 3 | * or more contributor license agreements. See the NOTICE file |
---|
| 4 | * distributed with this work for additional information |
---|
| 5 | * regarding copyright ownership. The ASF licenses this file |
---|
| 6 | * to you under the Apache License, Version 2.0 (the |
---|
| 7 | * "License"); you may not use this file except in compliance |
---|
| 8 | * with the License. You may obtain a copy of the License at |
---|
| 9 | * |
---|
| 10 | * http://www.apache.org/licenses/LICENSE-2.0 |
---|
| 11 | * |
---|
| 12 | * Unless required by applicable law or agreed to in writing, software |
---|
| 13 | * distributed under the License is distributed on an "AS IS" BASIS, |
---|
| 14 | * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
---|
| 15 | * See the License for the specific language governing permissions and |
---|
| 16 | * limitations under the License. |
---|
| 17 | */ |
---|
| 18 | |
---|
| 19 | package org.apache.hadoop.hdfs.server.namenode; |
---|
| 20 | |
---|
| 21 | import java.io.ByteArrayOutputStream; |
---|
| 22 | import java.io.PrintStream; |
---|
| 23 | import java.net.InetSocketAddress; |
---|
| 24 | import java.io.File; |
---|
| 25 | import java.io.RandomAccessFile; |
---|
| 26 | import java.lang.Exception; |
---|
| 27 | import java.io.IOException; |
---|
| 28 | import java.nio.channels.FileChannel; |
---|
| 29 | import java.util.Random; |
---|
| 30 | |
---|
| 31 | import junit.framework.TestCase; |
---|
| 32 | |
---|
| 33 | import org.apache.commons.logging.impl.Log4JLogger; |
---|
| 34 | import org.apache.log4j.Level; |
---|
| 35 | import org.apache.hadoop.conf.Configuration; |
---|
| 36 | import org.apache.hadoop.fs.FileSystem; |
---|
| 37 | import org.apache.hadoop.fs.Path; |
---|
| 38 | import org.apache.hadoop.fs.FSDataOutputStream; |
---|
| 39 | import org.apache.hadoop.util.ToolRunner; |
---|
| 40 | import org.apache.hadoop.hdfs.DFSClient; |
---|
| 41 | import org.apache.hadoop.hdfs.DFSTestUtil; |
---|
| 42 | import org.apache.hadoop.hdfs.MiniDFSCluster; |
---|
| 43 | import org.apache.hadoop.hdfs.protocol.LocatedBlocks; |
---|
| 44 | import org.apache.hadoop.hdfs.tools.DFSck; |
---|
| 45 | import org.apache.hadoop.io.IOUtils; |
---|
| 46 | |
---|
| 47 | /** |
---|
| 48 | * A JUnit test for doing fsck |
---|
| 49 | */ |
---|
| 50 | public class TestFsck extends TestCase { |
---|
| 51 | static String runFsck(Configuration conf, int expectedErrCode, |
---|
| 52 | boolean checkErrorCode,String... path) |
---|
| 53 | throws Exception { |
---|
| 54 | PrintStream oldOut = System.out; |
---|
| 55 | ByteArrayOutputStream bStream = new ByteArrayOutputStream(); |
---|
| 56 | PrintStream newOut = new PrintStream(bStream, true); |
---|
| 57 | System.setOut(newOut); |
---|
| 58 | ((Log4JLogger)PermissionChecker.LOG).getLogger().setLevel(Level.ALL); |
---|
| 59 | int errCode = ToolRunner.run(new DFSck(conf), path); |
---|
| 60 | if (checkErrorCode) |
---|
| 61 | assertEquals(expectedErrCode, errCode); |
---|
| 62 | ((Log4JLogger)PermissionChecker.LOG).getLogger().setLevel(Level.INFO); |
---|
| 63 | System.setOut(oldOut); |
---|
| 64 | return bStream.toString(); |
---|
| 65 | } |
---|
| 66 | |
---|
| 67 | /** do fsck */ |
---|
| 68 | public void testFsck() throws Exception { |
---|
| 69 | DFSTestUtil util = new DFSTestUtil("TestFsck", 20, 3, 8*1024); |
---|
| 70 | MiniDFSCluster cluster = null; |
---|
| 71 | FileSystem fs = null; |
---|
| 72 | try { |
---|
| 73 | Configuration conf = new Configuration(); |
---|
| 74 | conf.setLong("dfs.blockreport.intervalMsec", 10000L); |
---|
| 75 | cluster = new MiniDFSCluster(conf, 4, true, null); |
---|
| 76 | fs = cluster.getFileSystem(); |
---|
| 77 | util.createFiles(fs, "/srcdat"); |
---|
| 78 | util.waitReplication(fs, "/srcdat", (short)3); |
---|
| 79 | String outStr = runFsck(conf, 0, true, "/"); |
---|
| 80 | assertTrue(outStr.contains(NamenodeFsck.HEALTHY_STATUS)); |
---|
| 81 | System.out.println(outStr); |
---|
| 82 | if (fs != null) {try{fs.close();} catch(Exception e){}} |
---|
| 83 | cluster.shutdown(); |
---|
| 84 | |
---|
| 85 | // restart the cluster; bring up namenode but not the data nodes |
---|
| 86 | cluster = new MiniDFSCluster(conf, 0, false, null); |
---|
| 87 | outStr = runFsck(conf, 1, true, "/"); |
---|
| 88 | // expect the result is corrupt |
---|
| 89 | assertTrue(outStr.contains(NamenodeFsck.CORRUPT_STATUS)); |
---|
| 90 | System.out.println(outStr); |
---|
| 91 | |
---|
| 92 | // bring up data nodes & cleanup cluster |
---|
| 93 | cluster.startDataNodes(conf, 4, true, null, null); |
---|
| 94 | cluster.waitActive(); |
---|
| 95 | cluster.waitClusterUp(); |
---|
| 96 | fs = cluster.getFileSystem(); |
---|
| 97 | util.cleanup(fs, "/srcdat"); |
---|
| 98 | } finally { |
---|
| 99 | if (fs != null) {try{fs.close();} catch(Exception e){}} |
---|
| 100 | if (cluster != null) { cluster.shutdown(); } |
---|
| 101 | } |
---|
| 102 | } |
---|
| 103 | |
---|
| 104 | public void testFsckNonExistent() throws Exception { |
---|
| 105 | DFSTestUtil util = new DFSTestUtil("TestFsck", 20, 3, 8*1024); |
---|
| 106 | MiniDFSCluster cluster = null; |
---|
| 107 | FileSystem fs = null; |
---|
| 108 | try { |
---|
| 109 | Configuration conf = new Configuration(); |
---|
| 110 | conf.setLong("dfs.blockreport.intervalMsec", 10000L); |
---|
| 111 | cluster = new MiniDFSCluster(conf, 4, true, null); |
---|
| 112 | fs = cluster.getFileSystem(); |
---|
| 113 | util.createFiles(fs, "/srcdat"); |
---|
| 114 | util.waitReplication(fs, "/srcdat", (short)3); |
---|
| 115 | String outStr = runFsck(conf, 0, true, "/non-existent"); |
---|
| 116 | assertEquals(-1, outStr.indexOf(NamenodeFsck.HEALTHY_STATUS)); |
---|
| 117 | System.out.println(outStr); |
---|
| 118 | util.cleanup(fs, "/srcdat"); |
---|
| 119 | } finally { |
---|
| 120 | if (fs != null) {try{fs.close();} catch(Exception e){}} |
---|
| 121 | if (cluster != null) { cluster.shutdown(); } |
---|
| 122 | } |
---|
| 123 | } |
---|
| 124 | |
---|
| 125 | public void testFsckMove() throws Exception { |
---|
| 126 | DFSTestUtil util = new DFSTestUtil("TestFsck", 5, 3, 8*1024); |
---|
| 127 | MiniDFSCluster cluster = null; |
---|
| 128 | FileSystem fs = null; |
---|
| 129 | try { |
---|
| 130 | Configuration conf = new Configuration(); |
---|
| 131 | conf.setLong("dfs.blockreport.intervalMsec", 10000L); |
---|
| 132 | cluster = new MiniDFSCluster(conf, 4, true, null); |
---|
| 133 | String topDir = "/srcdat"; |
---|
| 134 | fs = cluster.getFileSystem(); |
---|
| 135 | cluster.waitActive(); |
---|
| 136 | util.createFiles(fs, topDir); |
---|
| 137 | util.waitReplication(fs, topDir, (short)3); |
---|
| 138 | String outStr = runFsck(conf, 0, true, "/"); |
---|
| 139 | assertTrue(outStr.contains(NamenodeFsck.HEALTHY_STATUS)); |
---|
| 140 | |
---|
| 141 | // Corrupt a block by deleting it |
---|
| 142 | String[] fileNames = util.getFileNames(topDir); |
---|
| 143 | DFSClient dfsClient = new DFSClient(new InetSocketAddress("localhost", |
---|
| 144 | cluster.getNameNodePort()), conf); |
---|
| 145 | String block = dfsClient.namenode. |
---|
| 146 | getBlockLocations(fileNames[0], 0, Long.MAX_VALUE). |
---|
| 147 | get(0).getBlock().getBlockName(); |
---|
| 148 | File baseDir = new File(System.getProperty("test.build.data", |
---|
| 149 | "build/test/data"),"dfs/data"); |
---|
| 150 | for (int i=0; i<8; i++) { |
---|
| 151 | File blockFile = new File(baseDir, "data" +(i+1)+ "/current/" + block); |
---|
| 152 | if(blockFile.exists()) { |
---|
| 153 | assertTrue(blockFile.delete()); |
---|
| 154 | } |
---|
| 155 | } |
---|
| 156 | |
---|
| 157 | // We excpect the filesystem to be corrupted |
---|
| 158 | outStr = runFsck(conf, 1, false, "/"); |
---|
| 159 | while (!outStr.contains(NamenodeFsck.CORRUPT_STATUS)) { |
---|
| 160 | try { |
---|
| 161 | Thread.sleep(100); |
---|
| 162 | } catch (InterruptedException ignore) { |
---|
| 163 | } |
---|
| 164 | outStr = runFsck(conf, 1, false, "/"); |
---|
| 165 | } |
---|
| 166 | |
---|
| 167 | // Fix the filesystem by moving corrupted files to lost+found |
---|
| 168 | outStr = runFsck(conf, 1, true, "/", "-move"); |
---|
| 169 | assertTrue(outStr.contains(NamenodeFsck.CORRUPT_STATUS)); |
---|
| 170 | |
---|
| 171 | // Check to make sure we have healthy filesystem |
---|
| 172 | outStr = runFsck(conf, 0, true, "/"); |
---|
| 173 | assertTrue(outStr.contains(NamenodeFsck.HEALTHY_STATUS)); |
---|
| 174 | util.cleanup(fs, topDir); |
---|
| 175 | if (fs != null) {try{fs.close();} catch(Exception e){}} |
---|
| 176 | cluster.shutdown(); |
---|
| 177 | } finally { |
---|
| 178 | if (fs != null) {try{fs.close();} catch(Exception e){}} |
---|
| 179 | if (cluster != null) { cluster.shutdown(); } |
---|
| 180 | } |
---|
| 181 | } |
---|
| 182 | |
---|
| 183 | public void testFsckOpenFiles() throws Exception { |
---|
| 184 | DFSTestUtil util = new DFSTestUtil("TestFsck", 4, 3, 8*1024); |
---|
| 185 | MiniDFSCluster cluster = null; |
---|
| 186 | FileSystem fs = null; |
---|
| 187 | try { |
---|
| 188 | Configuration conf = new Configuration(); |
---|
| 189 | conf.setLong("dfs.blockreport.intervalMsec", 10000L); |
---|
| 190 | cluster = new MiniDFSCluster(conf, 4, true, null); |
---|
| 191 | String topDir = "/srcdat"; |
---|
| 192 | String randomString = "HADOOP "; |
---|
| 193 | fs = cluster.getFileSystem(); |
---|
| 194 | cluster.waitActive(); |
---|
| 195 | util.createFiles(fs, topDir); |
---|
| 196 | util.waitReplication(fs, topDir, (short)3); |
---|
| 197 | String outStr = runFsck(conf, 0, true, "/"); |
---|
| 198 | assertTrue(outStr.contains(NamenodeFsck.HEALTHY_STATUS)); |
---|
| 199 | // Open a file for writing and do not close for now |
---|
| 200 | Path openFile = new Path(topDir + "/openFile"); |
---|
| 201 | FSDataOutputStream out = fs.create(openFile); |
---|
| 202 | int writeCount = 0; |
---|
| 203 | while (writeCount != 100) { |
---|
| 204 | out.write(randomString.getBytes()); |
---|
| 205 | writeCount++; |
---|
| 206 | } |
---|
| 207 | // We expect the filesystem to be HEALTHY and show one open file |
---|
| 208 | outStr = runFsck(conf, 0, true, topDir); |
---|
| 209 | System.out.println(outStr); |
---|
| 210 | assertTrue(outStr.contains(NamenodeFsck.HEALTHY_STATUS)); |
---|
| 211 | assertFalse(outStr.contains("OPENFORWRITE")); |
---|
| 212 | // Use -openforwrite option to list open files |
---|
| 213 | outStr = runFsck(conf, 0, true, topDir, "-openforwrite"); |
---|
| 214 | System.out.println(outStr); |
---|
| 215 | assertTrue(outStr.contains("OPENFORWRITE")); |
---|
| 216 | assertTrue(outStr.contains("openFile")); |
---|
| 217 | // Close the file |
---|
| 218 | out.close(); |
---|
| 219 | // Now, fsck should show HEALTHY fs and should not show any open files |
---|
| 220 | outStr = runFsck(conf, 0, true, topDir); |
---|
| 221 | System.out.println(outStr); |
---|
| 222 | assertTrue(outStr.contains(NamenodeFsck.HEALTHY_STATUS)); |
---|
| 223 | assertFalse(outStr.contains("OPENFORWRITE")); |
---|
| 224 | util.cleanup(fs, topDir); |
---|
| 225 | if (fs != null) {try{fs.close();} catch(Exception e){}} |
---|
| 226 | cluster.shutdown(); |
---|
| 227 | } finally { |
---|
| 228 | if (fs != null) {try{fs.close();} catch(Exception e){}} |
---|
| 229 | if (cluster != null) { cluster.shutdown(); } |
---|
| 230 | } |
---|
| 231 | } |
---|
| 232 | |
---|
| 233 | public void testCorruptBlock() throws Exception { |
---|
| 234 | Configuration conf = new Configuration(); |
---|
| 235 | conf.setLong("dfs.blockreport.intervalMsec", 1000); |
---|
| 236 | FileSystem fs = null; |
---|
| 237 | DFSClient dfsClient = null; |
---|
| 238 | LocatedBlocks blocks = null; |
---|
| 239 | int replicaCount = 0; |
---|
| 240 | Random random = new Random(); |
---|
| 241 | String outStr = null; |
---|
| 242 | |
---|
| 243 | MiniDFSCluster cluster = null; |
---|
| 244 | try { |
---|
| 245 | cluster = new MiniDFSCluster(conf, 3, true, null); |
---|
| 246 | cluster.waitActive(); |
---|
| 247 | fs = cluster.getFileSystem(); |
---|
| 248 | Path file1 = new Path("/testCorruptBlock"); |
---|
| 249 | DFSTestUtil.createFile(fs, file1, 1024, (short)3, 0); |
---|
| 250 | // Wait until file replication has completed |
---|
| 251 | DFSTestUtil.waitReplication(fs, file1, (short)3); |
---|
| 252 | String block = DFSTestUtil.getFirstBlock(fs, file1).getBlockName(); |
---|
| 253 | |
---|
| 254 | // Make sure filesystem is in healthy state |
---|
| 255 | outStr = runFsck(conf, 0, true, "/"); |
---|
| 256 | System.out.println(outStr); |
---|
| 257 | assertTrue(outStr.contains(NamenodeFsck.HEALTHY_STATUS)); |
---|
| 258 | |
---|
| 259 | // corrupt replicas |
---|
| 260 | File baseDir = new File(System.getProperty("test.build.data", |
---|
| 261 | "build/test/data"),"dfs/data"); |
---|
| 262 | for (int i=0; i < 6; i++) { |
---|
| 263 | File blockFile = new File(baseDir, "data" + (i+1) + "/current/" + |
---|
| 264 | block); |
---|
| 265 | if (blockFile.exists()) { |
---|
| 266 | RandomAccessFile raFile = new RandomAccessFile(blockFile, "rw"); |
---|
| 267 | FileChannel channel = raFile.getChannel(); |
---|
| 268 | String badString = "BADBAD"; |
---|
| 269 | int rand = random.nextInt((int)channel.size()/2); |
---|
| 270 | raFile.seek(rand); |
---|
| 271 | raFile.write(badString.getBytes()); |
---|
| 272 | raFile.close(); |
---|
| 273 | } |
---|
| 274 | } |
---|
| 275 | // Read the file to trigger reportBadBlocks |
---|
| 276 | try { |
---|
| 277 | IOUtils.copyBytes(fs.open(file1), new IOUtils.NullOutputStream(), conf, |
---|
| 278 | true); |
---|
| 279 | } catch (IOException ie) { |
---|
| 280 | // Ignore exception |
---|
| 281 | } |
---|
| 282 | |
---|
| 283 | dfsClient = new DFSClient(new InetSocketAddress("localhost", |
---|
| 284 | cluster.getNameNodePort()), conf); |
---|
| 285 | blocks = dfsClient.namenode. |
---|
| 286 | getBlockLocations(file1.toString(), 0, Long.MAX_VALUE); |
---|
| 287 | replicaCount = blocks.get(0).getLocations().length; |
---|
| 288 | while (replicaCount != 3) { |
---|
| 289 | try { |
---|
| 290 | Thread.sleep(100); |
---|
| 291 | } catch (InterruptedException ignore) { |
---|
| 292 | } |
---|
| 293 | blocks = dfsClient.namenode. |
---|
| 294 | getBlockLocations(file1.toString(), 0, Long.MAX_VALUE); |
---|
| 295 | replicaCount = blocks.get(0).getLocations().length; |
---|
| 296 | } |
---|
| 297 | assertTrue (blocks.get(0).isCorrupt()); |
---|
| 298 | |
---|
| 299 | // Check if fsck reports the same |
---|
| 300 | outStr = runFsck(conf, 1, true, "/"); |
---|
| 301 | System.out.println(outStr); |
---|
| 302 | assertTrue(outStr.contains(NamenodeFsck.CORRUPT_STATUS)); |
---|
| 303 | assertTrue(outStr.contains("testCorruptBlock")); |
---|
| 304 | } finally { |
---|
| 305 | if (cluster != null) {cluster.shutdown();} |
---|
| 306 | } |
---|
| 307 | } |
---|
| 308 | |
---|
| 309 | /** Test if fsck can return -1 in case of failure |
---|
| 310 | * |
---|
| 311 | * @throws Exception |
---|
| 312 | */ |
---|
| 313 | public void testFsckError() throws Exception { |
---|
| 314 | MiniDFSCluster cluster = null; |
---|
| 315 | try { |
---|
| 316 | // bring up a one-node cluster |
---|
| 317 | Configuration conf = new Configuration(); |
---|
| 318 | cluster = new MiniDFSCluster(conf, 1, true, null); |
---|
| 319 | String fileName = "/test.txt"; |
---|
| 320 | Path filePath = new Path(fileName); |
---|
| 321 | FileSystem fs = cluster.getFileSystem(); |
---|
| 322 | |
---|
| 323 | // create a one-block file |
---|
| 324 | DFSTestUtil.createFile(fs, filePath, 1L, (short)1, 1L); |
---|
| 325 | DFSTestUtil.waitReplication(fs, filePath, (short)1); |
---|
| 326 | |
---|
| 327 | // intentionally corrupt NN data structure |
---|
| 328 | INodeFile node = (INodeFile)cluster.getNameNode().namesystem.dir.rootDir.getNode(fileName); |
---|
| 329 | assertEquals(node.blocks.length, 1); |
---|
| 330 | node.blocks[0].setNumBytes(-1L); // set the block length to be negative |
---|
| 331 | |
---|
| 332 | // run fsck and expect a failure with -1 as the error code |
---|
| 333 | String outStr = runFsck(conf, -1, true, fileName); |
---|
| 334 | System.out.println(outStr); |
---|
| 335 | assertTrue(outStr.contains(NamenodeFsck.FAILURE_STATUS)); |
---|
| 336 | |
---|
| 337 | // clean up file system |
---|
| 338 | fs.delete(filePath, true); |
---|
| 339 | } finally { |
---|
| 340 | if (cluster != null) {cluster.shutdown();} |
---|
| 341 | } |
---|
| 342 | } |
---|
| 343 | } |
---|