Repository: hbase Updated Branches: refs/heads/0.98 be48e80d4 -> f73f18abb
HBASE-11405 Multiple invocations of hbck in parallel disables balancer permanently (Sean Busbey and bharath v) Project: http://git-wip-us.apache.org/repos/asf/hbase/repo Commit: http://git-wip-us.apache.org/repos/asf/hbase/commit/f73f18ab Tree: http://git-wip-us.apache.org/repos/asf/hbase/tree/f73f18ab Diff: http://git-wip-us.apache.org/repos/asf/hbase/diff/f73f18ab Branch: refs/heads/0.98 Commit: f73f18abbb26e37e955e77e1a7698d1a50aaa32c Parents: be48e80 Author: Ted Yu <[email protected]> Authored: Thu Sep 25 16:28:17 2014 +0000 Committer: Ted Yu <[email protected]> Committed: Thu Sep 25 16:28:17 2014 +0000 ---------------------------------------------------------------------- .../org/apache/hadoop/hbase/util/HBaseFsck.java | 82 +++++++++++++++++++- .../hadoop/hbase/HBaseTestingUtility.java | 58 ++++++++++++++ .../apache/hadoop/hbase/util/TestHBaseFsck.java | 50 ++++++++++++ 3 files changed, 189 insertions(+), 1 deletion(-) ---------------------------------------------------------------------- http://git-wip-us.apache.org/repos/asf/hbase/blob/f73f18ab/hbase-server/src/main/java/org/apache/hadoop/hbase/util/HBaseFsck.java ---------------------------------------------------------------------- diff --git a/hbase-server/src/main/java/org/apache/hadoop/hbase/util/HBaseFsck.java b/hbase-server/src/main/java/org/apache/hadoop/hbase/util/HBaseFsck.java index edd8b66..d0a19ba 100644 --- a/hbase-server/src/main/java/org/apache/hadoop/hbase/util/HBaseFsck.java +++ b/hbase-server/src/main/java/org/apache/hadoop/hbase/util/HBaseFsck.java @@ -21,6 +21,7 @@ import java.io.FileNotFoundException; import java.io.IOException; import java.io.PrintWriter; import java.io.StringWriter; +import java.net.InetAddress; import java.net.URI; import java.util.ArrayList; import java.util.Arrays; @@ -45,6 +46,7 @@ import java.util.concurrent.ExecutorService; import java.util.concurrent.Future; import java.util.concurrent.ScheduledThreadPoolExecutor; import java.util.concurrent.atomic.AtomicInteger; +import java.util.concurrent.atomic.AtomicBoolean; import org.apache.commons.logging.Log; import org.apache.commons.logging.LogFactory; @@ -52,10 +54,12 @@ import org.apache.hadoop.hbase.classification.InterfaceAudience; import org.apache.hadoop.hbase.classification.InterfaceStability; import org.apache.hadoop.conf.Configuration; import org.apache.hadoop.conf.Configured; +import org.apache.hadoop.fs.FSDataOutputStream; import org.apache.hadoop.fs.FileStatus; import org.apache.hadoop.fs.FileSystem; import org.apache.hadoop.fs.Path; import org.apache.hadoop.fs.permission.FsAction; +import org.apache.hadoop.fs.permission.FsPermission; import org.apache.hadoop.hbase.Abortable; import org.apache.hadoop.hbase.Cell; import org.apache.hadoop.hbase.ClusterStatus; @@ -105,7 +109,10 @@ import org.apache.hadoop.hbase.zookeeper.MetaRegionTracker; import org.apache.hadoop.hbase.zookeeper.ZKTableReadOnly; import org.apache.hadoop.hbase.zookeeper.ZooKeeperWatcher; import org.apache.hadoop.hbase.security.AccessDeniedException; +import org.apache.hadoop.hdfs.protocol.AlreadyBeingCreatedException; import org.apache.hadoop.io.IOUtils; +import org.apache.hadoop.ipc.RemoteException; +import org.apache.hadoop.security.AccessControlException; import org.apache.hadoop.security.UserGroupInformation; import org.apache.hadoop.util.ReflectionUtils; import org.apache.hadoop.util.Tool; @@ -174,6 +181,8 @@ public class HBaseFsck extends Configured { private static final int DEFAULT_OVERLAPS_TO_SIDELINE = 2; private static final int DEFAULT_MAX_MERGE = 5; private static final String TO_BE_LOADED = "to_be_loaded"; + private static final String HBCK_LOCK_FILE = "hbase-hbck.lock"; + /********************** * Internal resources @@ -188,6 +197,12 @@ public class HBaseFsck extends Configured { private long startMillis = System.currentTimeMillis(); private HFileCorruptionChecker hfcc; private int retcode = 0; + private Path HBCK_LOCK_PATH; + private FSDataOutputStream hbckOutFd; + // This lock is to prevent cleanup of balancer resources twice between + // ShutdownHook and the main code. We cleanup only if the connect() is + // successful + private final AtomicBoolean hbckLockCleanup = new AtomicBoolean(false); /*********** * Options @@ -297,10 +312,73 @@ public class HBaseFsck extends Configured { } /** + * This method maintains a lock using a file. If the creation fails we return null + * + * @return FSDataOutputStream object corresponding to the newly opened lock file + * @throws IOException + */ + private FSDataOutputStream checkAndMarkRunningHbck() throws IOException { + try { + FileSystem fs = FSUtils.getCurrentFileSystem(getConf()); + FsPermission defaultPerms = FSUtils.getFilePermissions(fs, getConf(), + HConstants.DATA_FILE_UMASK_KEY); + Path tmpDir = new Path(FSUtils.getRootDir(getConf()), HConstants.HBASE_TEMP_DIRECTORY); + fs.mkdirs(tmpDir); + HBCK_LOCK_PATH = new Path(tmpDir, HBCK_LOCK_FILE); + final FSDataOutputStream out = FSUtils.create(fs, HBCK_LOCK_PATH, defaultPerms, false); + out.writeBytes(InetAddress.getLocalHost().toString()); + out.flush(); + return out; + } catch(RemoteException e) { + if(AlreadyBeingCreatedException.class.getName().equals(e.getClassName())){ + return null; + } else { + throw e; + } + } + } + + private void unlockHbck() { + if(hbckLockCleanup.compareAndSet(true, false)){ + IOUtils.closeStream(hbckOutFd); + try{ + FSUtils.delete(FSUtils.getCurrentFileSystem(getConf()), HBCK_LOCK_PATH, true); + } catch(IOException ioe) { + LOG.warn("Failed to delete " + HBCK_LOCK_PATH); + LOG.debug(ioe); + } + } + } + + /** * To repair region consistency, one must call connect() in order to repair * online state. */ public void connect() throws IOException { + + // Check if another instance of balancer is running + hbckOutFd = checkAndMarkRunningHbck(); + if (hbckOutFd == null) { + setRetCode(-1); + LOG.error("Another instance of hbck is running, exiting this instance.[If you are sure" + + " no other instance is running, delete the lock file " + + HBCK_LOCK_PATH + " and rerun the tool]"); + throw new IOException("Duplicate hbck - Abort"); + } + + // Make sure to cleanup the lock + hbckLockCleanup.set(true); + + // Add a shutdown hook to this thread, incase user tries to + // kill the hbck with a ctrl-c, we want to cleanup the lock so that + // it is available for further calls + Runtime.getRuntime().addShutdownHook(new Thread() { + public void run() { + unlockHbck(); + } + }); + LOG.debug("Launching hbck"); + connection = HConnectionManager.createConnection(getConf()); admin = new HBaseAdmin(connection); meta = new HTable(TableName.META_TABLE_NAME, connection); @@ -495,6 +573,9 @@ public class HBaseFsck extends Configured { checkAndFixTableLocks(); + // Remove the hbck lock + unlockHbck(); + // Print table summary printTableSummary(tablesInfo); return errors.summarize(); @@ -3825,7 +3906,6 @@ public class HBaseFsck extends Configured { Path hbasedir = FSUtils.getRootDir(conf); URI defaultFs = hbasedir.getFileSystem(conf).getUri(); FSUtils.setFsDefault(conf, new Path(defaultFs)); - int ret = ToolRunner.run(new HBaseFsckTool(conf), args); System.exit(ret); } http://git-wip-us.apache.org/repos/asf/hbase/blob/f73f18ab/hbase-server/src/test/java/org/apache/hadoop/hbase/HBaseTestingUtility.java ---------------------------------------------------------------------- diff --git a/hbase-server/src/test/java/org/apache/hadoop/hbase/HBaseTestingUtility.java b/hbase-server/src/test/java/org/apache/hadoop/hbase/HBaseTestingUtility.java index d4f4888..b93e565 100644 --- a/hbase-server/src/test/java/org/apache/hadoop/hbase/HBaseTestingUtility.java +++ b/hbase-server/src/test/java/org/apache/hadoop/hbase/HBaseTestingUtility.java @@ -27,6 +27,7 @@ import java.lang.reflect.Field; import java.lang.reflect.Method; import java.lang.reflect.Modifier; import java.net.InetAddress; +import java.net.InetSocketAddress; import java.net.ServerSocket; import java.net.Socket; import java.net.UnknownHostException; @@ -103,6 +104,7 @@ import org.apache.hadoop.hdfs.MiniDFSCluster; import org.apache.hadoop.mapred.JobConf; import org.apache.hadoop.mapred.MiniMRCluster; import org.apache.hadoop.mapred.TaskLog; +import org.apache.hadoop.security.UserGroupInformation; import org.apache.zookeeper.KeeperException; import org.apache.zookeeper.KeeperException.NodeExistsException; import org.apache.zookeeper.WatchedEvent; @@ -262,6 +264,18 @@ public class HBaseTestingUtility extends HBaseCommonTestingUtility { } /** + * Controls how many attempts we will make in the face of failures in HDFS. + * @deprecated to be removed with Hadoop 1.x support + */ + @Deprecated + public void setHDFSClientRetry(final int retries) { + this.conf.setInt("hdfs.client.retries.number", retries); + if (0 == retries) { + makeDFSClientNonRetrying(); + } + } + + /** * Returns this classes's instance of {@link Configuration}. Be careful how * you use the returned Configuration since {@link HConnection} instances * can be shared. The Map of HConnections is keyed by the Configuration. If @@ -1927,6 +1941,50 @@ public class HBaseTestingUtility extends HBaseCommonTestingUtility { return createMultiRegions(c, table, columnFamily, KEYS); } + void makeDFSClientNonRetrying() { + if (null == this.dfsCluster) { + LOG.debug("dfsCluster has not started, can't make client non-retrying."); + return; + } + try { + final FileSystem filesystem = this.dfsCluster.getFileSystem(); + if (!(filesystem instanceof DistributedFileSystem)) { + LOG.debug("dfsCluster is not backed by a DistributedFileSystem, can't make client non-retrying."); + return; + } + // rely on FileSystem.CACHE to alter how we talk via DFSClient + final DistributedFileSystem fs = (DistributedFileSystem)filesystem; + // retrieve the backing DFSClient instance + final Field dfsField = fs.getClass().getDeclaredField("dfs"); + dfsField.setAccessible(true); + final Class<?> dfsClazz = dfsField.getType(); + final DFSClient dfs = DFSClient.class.cast(dfsField.get(fs)); + + // expose the method for creating direct RPC connections. + final Method createRPCNamenode = dfsClazz.getDeclaredMethod("createRPCNamenode", InetSocketAddress.class, Configuration.class, UserGroupInformation.class); + createRPCNamenode.setAccessible(true); + + // grab the DFSClient instance's backing connection information + final Field nnField = dfsClazz.getDeclaredField("nnAddress"); + nnField.setAccessible(true); + final InetSocketAddress nnAddress = InetSocketAddress.class.cast(nnField.get(dfs)); + final Field confField = dfsClazz.getDeclaredField("conf"); + confField.setAccessible(true); + final Configuration conf = Configuration.class.cast(confField.get(dfs)); + final Field ugiField = dfsClazz.getDeclaredField("ugi"); + ugiField.setAccessible(true); + final UserGroupInformation ugi = UserGroupInformation.class.cast(ugiField.get(dfs)); + + // replace the proxy for the namenode rpc with a direct instance + final Field namenodeField = dfsClazz.getDeclaredField("namenode"); + namenodeField.setAccessible(true); + namenodeField.set(dfs, createRPCNamenode.invoke(null, nnAddress, conf, ugi)); + LOG.debug("Set DSFClient namenode to bare RPC"); + } catch (Exception exception) { + LOG.info("Could not alter DFSClient to be non-retrying.", exception); + } + } + /** * Creates the specified number of regions in the specified table. * @param c http://git-wip-us.apache.org/repos/asf/hbase/blob/f73f18ab/hbase-server/src/test/java/org/apache/hadoop/hbase/util/TestHBaseFsck.java ---------------------------------------------------------------------- diff --git a/hbase-server/src/test/java/org/apache/hadoop/hbase/util/TestHBaseFsck.java b/hbase-server/src/test/java/org/apache/hadoop/hbase/util/TestHBaseFsck.java index 62149e5..c903c12 100644 --- a/hbase-server/src/test/java/org/apache/hadoop/hbase/util/TestHBaseFsck.java +++ b/hbase-server/src/test/java/org/apache/hadoop/hbase/util/TestHBaseFsck.java @@ -36,8 +36,11 @@ import java.util.LinkedList; import java.util.List; import java.util.Map; import java.util.Map.Entry; +import java.util.concurrent.Callable; import java.util.concurrent.CountDownLatch; import java.util.concurrent.ExecutorService; +import java.util.concurrent.Executors; +import java.util.concurrent.Future; import java.util.concurrent.ScheduledThreadPoolExecutor; import java.util.concurrent.SynchronousQueue; import java.util.concurrent.ThreadPoolExecutor; @@ -127,11 +130,13 @@ public class TestHBaseFsck { Bytes.toBytes("00"), Bytes.toBytes("50"), Bytes.toBytes("A0"), Bytes.toBytes("A5"), Bytes.toBytes("B0"), Bytes.toBytes("B5"), Bytes.toBytes("C0"), Bytes.toBytes("C5") }; + @SuppressWarnings("deprecation") @BeforeClass public static void setUpBeforeClass() throws Exception { TEST_UTIL.getConfiguration().setInt("hbase.regionserver.handler.count", 2); TEST_UTIL.getConfiguration().setInt("hbase.regionserver.metahandler.count", 2); TEST_UTIL.startMiniCluster(3); + TEST_UTIL.setHDFSClientRetry(0); executorService = new ThreadPoolExecutor(1, Integer.MAX_VALUE, 60, TimeUnit.SECONDS, new SynchronousQueue<Runnable>(), Threads.newDaemonThreadFactory("testhbck")); @@ -513,6 +518,51 @@ public class TestHBaseFsck { } /** + * This test makes sure that parallel instances of Hbck is disabled. + * + * @throws Exception + */ + @Test + public void testParallelHbck() throws Exception { + final ExecutorService service; + final Future<HBaseFsck> hbck1,hbck2; + + class RunHbck implements Callable<HBaseFsck>{ + boolean fail = true; + public HBaseFsck call(){ + try{ + return doFsck(conf, false); + } catch(Exception e){ + if (e.getMessage().contains("Duplicate hbck")) { + fail = false; + } else { + LOG.fatal("hbck failed.", e); + } + } + // If we reach here, then an exception was caught + if (fail) fail(); + return null; + } + } + service = Executors.newFixedThreadPool(2); + hbck1 = service.submit(new RunHbck()); + hbck2 = service.submit(new RunHbck()); + service.shutdown(); + //wait for 15 seconds, for both hbck calls finish + service.awaitTermination(15, TimeUnit.SECONDS); + HBaseFsck h1 = hbck1.get(); + HBaseFsck h2 = hbck2.get(); + // Make sure only one of the calls was successful + assert(h1 == null || h2 == null); + if (h1 != null) { + assert(h1.getRetCode() >= 0); + } + if (h2 != null) { + assert(h2.getRetCode() >= 0); + } + } + + /** * This create and fixes a bad table with regions that have a duplicate * start key */
