hbase-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From jmhs...@apache.org
Subject svn commit: r1446147 [12/35] - in /hbase/branches/hbase-7290v2: ./ bin/ conf/ dev-support/ hbase-client/ hbase-common/ hbase-common/src/main/java/org/apache/hadoop/hbase/ hbase-common/src/main/java/org/apache/hadoop/hbase/io/compress/ hbase-common/src/...
Date Thu, 14 Feb 2013 12:58:21 GMT
Modified: hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/HMaster.java
URL: http://svn.apache.org/viewvc/hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/HMaster.java?rev=1446147&r1=1446146&r2=1446147&view=diff
==============================================================================
--- hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/HMaster.java (original)
+++ hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/HMaster.java Thu Feb 14 12:58:12 2013
@@ -54,6 +54,7 @@ import org.apache.hadoop.hbase.HColumnDe
 import org.apache.hadoop.hbase.HConstants;
 import org.apache.hadoop.hbase.HRegionInfo;
 import org.apache.hadoop.hbase.HTableDescriptor;
+import org.apache.hadoop.hbase.HealthCheckChore;
 import org.apache.hadoop.hbase.MasterAdminProtocol;
 import org.apache.hadoop.hbase.MasterMonitorProtocol;
 import org.apache.hadoop.hbase.MasterNotRunningException;
@@ -78,11 +79,11 @@ import org.apache.hadoop.hbase.client.Re
 import org.apache.hadoop.hbase.coprocessor.CoprocessorHost;
 import org.apache.hadoop.hbase.executor.ExecutorService;
 import org.apache.hadoop.hbase.executor.ExecutorService.ExecutorType;
-import org.apache.hadoop.hbase.ipc.HBaseRPC;
 import org.apache.hadoop.hbase.ipc.HBaseServer;
-import org.apache.hadoop.hbase.ipc.ProtocolSignature;
+import org.apache.hadoop.hbase.ipc.HBaseServerRPC;
 import org.apache.hadoop.hbase.ipc.RpcServer;
 import org.apache.hadoop.hbase.ipc.ServerRpcController;
+import org.apache.hadoop.hbase.ipc.UnknownProtocolException;
 import org.apache.hadoop.hbase.master.balancer.BalancerChore;
 import org.apache.hadoop.hbase.master.balancer.ClusterStatusChore;
 import org.apache.hadoop.hbase.master.balancer.LoadBalancerFactory;
@@ -96,7 +97,6 @@ import org.apache.hadoop.hbase.master.ha
 import org.apache.hadoop.hbase.master.handler.ServerShutdownHandler;
 import org.apache.hadoop.hbase.master.handler.TableAddFamilyHandler;
 import org.apache.hadoop.hbase.master.handler.TableDeleteFamilyHandler;
-import org.apache.hadoop.hbase.master.handler.TableEventHandler;
 import org.apache.hadoop.hbase.master.handler.TableModifyFamilyHandler;
 import org.apache.hadoop.hbase.master.snapshot.SnapshotManager;
 import org.apache.hadoop.hbase.monitoring.MemoryBoundedLogMessageBuffer;
@@ -337,6 +337,9 @@ Server {
   // monitor for snapshot of hbase tables
   private SnapshotManager snapshotManager;
 
+  /** The health check chore. */
+  private HealthCheckChore healthCheckChore;
+
   /**
    * Initializes the HMaster. The steps are as follows:
    * <p>
@@ -368,9 +371,9 @@ Server {
     }
     int numHandlers = conf.getInt("hbase.master.handler.count",
       conf.getInt("hbase.regionserver.handler.count", 25));
-    this.rpcServer = HBaseRPC.getServer(MasterMonitorProtocol.class, this,
-      new Class<?>[]{MasterMonitorProtocol.class,
-        MasterAdminProtocol.class, RegionServerStatusProtocol.class},
+    this.rpcServer = HBaseServerRPC.getServer(MasterMonitorProtocol.class, this,
+        new Class<?>[]{MasterMonitorProtocol.class,
+            MasterAdminProtocol.class, RegionServerStatusProtocol.class},
         initialIsa.getHostName(), // BindAddress is IP we got for this server.
         initialIsa.getPort(),
         numHandlers,
@@ -416,6 +419,13 @@ Server {
     this.masterCheckCompression = conf.getBoolean("hbase.master.check.compression", true);
 
     this.metricsMaster = new MetricsMaster( new MetricsMasterWrapperImpl(this));
+
+    // Health checker thread.
+    int sleepTime = this.conf.getInt(HConstants.HEALTH_CHORE_WAKE_FREQ,
+      HConstants.DEFAULT_THREAD_WAKE_FREQUENCY);
+    if (isHealthCheckerConfigured()) {
+      healthCheckChore = new HealthCheckChore(sleepTime, this, getConfiguration());
+    }
   }
 
   /**
@@ -440,7 +450,8 @@ Server {
     while (!amm.isActiveMaster()) {
       LOG.debug("Waiting for master address ZNode to be written " +
         "(Also watching cluster state node)");
-      Thread.sleep(c.getInt("zookeeper.session.timeout", 180 * 1000));
+      Thread.sleep(
+        c.getInt(HConstants.ZK_SESSION_TIMEOUT, HConstants.DEFAULT_ZK_SESSION_TIMEOUT));
     }
 
   }
@@ -548,8 +559,7 @@ Server {
    */
   private void initializeZKBasedSystemTrackers() throws IOException,
       InterruptedException, KeeperException {
-    this.catalogTracker = createCatalogTracker(this.zooKeeper, this.conf,
-        this, conf.getInt("hbase.master.catalog.timeout", 600000));
+    this.catalogTracker = createCatalogTracker(this.zooKeeper, this.conf, this);
     this.catalogTracker.start();
 
     this.balancer = LoadBalancerFactory.getLoadBalancer(conf);
@@ -595,9 +605,9 @@ Server {
    * @throws IOException
    */
   CatalogTracker createCatalogTracker(final ZooKeeperWatcher zk,
-      final Configuration conf, Abortable abortable, final int defaultTimeout)
+      final Configuration conf, Abortable abortable)
   throws IOException {
-    return new CatalogTracker(zk, conf, abortable, defaultTimeout);
+    return new CatalogTracker(zk, conf, abortable);
   }
 
   // Check if we should stop every 100ms
@@ -970,38 +980,12 @@ Server {
       return;
     }
     LOG.info("Forcing splitLog and expire of " + sn);
+    fileSystemManager.splitMetaLog(sn);
     fileSystemManager.splitLog(sn);
     serverManager.expireServer(sn);
   }
 
   @Override
-  public ProtocolSignature getProtocolSignature(
-      String protocol, long version, int clientMethodsHashCode)
-  throws IOException {
-    if (MasterMonitorProtocol.class.getName().equals(protocol)) {
-      return new ProtocolSignature(MasterMonitorProtocol.VERSION, null);
-    } else if (MasterAdminProtocol.class.getName().equals(protocol)) {
-      return new ProtocolSignature(MasterAdminProtocol.VERSION, null);
-    } else if (RegionServerStatusProtocol.class.getName().equals(protocol)) {
-      return new ProtocolSignature(RegionServerStatusProtocol.VERSION, null);
-    }
-    throw new IOException("Unknown protocol: " + protocol);
-  }
-
-  public long getProtocolVersion(String protocol, long clientVersion) {
-    if (MasterMonitorProtocol.class.getName().equals(protocol)) {
-      return MasterMonitorProtocol.VERSION;
-    } else if (MasterAdminProtocol.class.getName().equals(protocol)) {
-      return MasterAdminProtocol.VERSION;
-    } else if (RegionServerStatusProtocol.class.getName().equals(protocol)) {
-      return RegionServerStatusProtocol.VERSION;
-    }
-    // unknown protocol
-    LOG.warn("Version requested for unimplemented protocol: "+protocol);
-    return -1;
-  }
-
-  @Override
   public TableDescriptors getTableDescriptors() {
     return this.tableDescriptors;
   }
@@ -1089,6 +1073,11 @@ Server {
      this.infoServer.start();
     }
 
+   // Start the health checker
+   if (this.healthCheckChore != null) {
+     Threads.setDaemonThreadRunning(this.healthCheckChore.getThread(), n + ".healthChecker");
+   }
+
     // Start allowing requests to happen.
     this.rpcServer.openServer();
     this.rpcServerOpen = true;
@@ -1124,6 +1113,9 @@ Server {
       }
     }
     if (this.executorService != null) this.executorService.shutdown();
+    if (this.healthCheckChore != null) {
+      this.healthCheckChore.interrupt();
+    }
   }
 
   private static Thread getAndStartClusterStatusChore(HMaster master) {
@@ -1579,19 +1571,22 @@ Server {
   }
 
   @Override
+  public void deleteTable(final byte[] tableName) throws IOException {
+    checkInitialized();
+    if (cpHost != null) {
+      cpHost.preDeleteTable(tableName);
+    }
+    this.executorService.submit(new DeleteTableHandler(tableName, this, this));
+    if (cpHost != null) {
+      cpHost.postDeleteTable(tableName);
+    }
+  }
+
+  @Override
   public DeleteTableResponse deleteTable(RpcController controller, DeleteTableRequest request)
   throws ServiceException {
-    byte [] tableName = request.getTableName().toByteArray();
     try {
-      checkInitialized();
-      if (cpHost != null) {
-        cpHost.preDeleteTable(tableName);
-      }
-      this.executorService.submit(new DeleteTableHandler(tableName, this, this));
-
-      if (cpHost != null) {
-        cpHost.postDeleteTable(tableName);
-      }
+      deleteTable(request.getTableName().toByteArray());
     } catch (IOException ioe) {
       throw new ServiceException(ioe);
     }
@@ -1626,45 +1621,55 @@ Server {
     }
   }
 
+  @Override
+  public void addColumn(final byte[] tableName, final HColumnDescriptor column)
+      throws IOException {
+    checkInitialized();
+    if (cpHost != null) {
+      if (cpHost.preAddColumn(tableName, column)) {
+        return;
+      }
+    }
+    new TableAddFamilyHandler(tableName, column, this, this).process();
+    if (cpHost != null) {
+      cpHost.postAddColumn(tableName, column);
+    }
+  }
+
+  @Override
   public AddColumnResponse addColumn(RpcController controller, AddColumnRequest req)
   throws ServiceException {
-    byte [] tableName = req.getTableName().toByteArray();
-    HColumnDescriptor column = HColumnDescriptor.convert(req.getColumnFamilies());
-
     try {
-      checkInitialized();
-      if (cpHost != null) {
-        if (cpHost.preAddColumn(tableName, column)) {
-          return AddColumnResponse.newBuilder().build();
-        }
-      }
-      new TableAddFamilyHandler(tableName, column, this, this).process();
-      if (cpHost != null) {
-        cpHost.postAddColumn(tableName, column);
-      }
+      addColumn(req.getTableName().toByteArray(),
+        HColumnDescriptor.convert(req.getColumnFamilies()));
     } catch (IOException ioe) {
       throw new ServiceException(ioe);
     }
     return AddColumnResponse.newBuilder().build();
   }
 
+  @Override
+  public void modifyColumn(byte[] tableName, HColumnDescriptor descriptor)
+      throws IOException {
+    checkInitialized();
+    checkCompression(descriptor);
+    if (cpHost != null) {
+      if (cpHost.preModifyColumn(tableName, descriptor)) {
+        return;
+      }
+    }
+    new TableModifyFamilyHandler(tableName, descriptor, this, this).process();
+    if (cpHost != null) {
+      cpHost.postModifyColumn(tableName, descriptor);
+    }
+  }
+
+  @Override
   public ModifyColumnResponse modifyColumn(RpcController controller, ModifyColumnRequest req)
   throws ServiceException {
-    byte [] tableName = req.getTableName().toByteArray();
-    HColumnDescriptor descriptor = HColumnDescriptor.convert(req.getColumnFamilies());
-
     try {
-      checkInitialized();
-      checkCompression(descriptor);
-      if (cpHost != null) {
-        if (cpHost.preModifyColumn(tableName, descriptor)) {
-          return ModifyColumnResponse.newBuilder().build();
-        }
-      }
-      new TableModifyFamilyHandler(tableName, descriptor, this, this).process();
-      if (cpHost != null) {
-        cpHost.postModifyColumn(tableName, descriptor);
-      }
+      modifyColumn(req.getTableName().toByteArray(),
+        HColumnDescriptor.convert(req.getColumnFamilies()));
     } catch (IOException ioe) {
       throw new ServiceException(ioe);
     }
@@ -1672,21 +1677,25 @@ Server {
   }
 
   @Override
+  public void deleteColumn(final byte[] tableName, final byte[] columnName)
+      throws IOException {
+    checkInitialized();
+    if (cpHost != null) {
+      if (cpHost.preDeleteColumn(tableName, columnName)) {
+        return;
+      }
+    }
+    new TableDeleteFamilyHandler(tableName, columnName, this, this).process();
+    if (cpHost != null) {
+      cpHost.postDeleteColumn(tableName, columnName);
+    }
+  }
+
+  @Override
   public DeleteColumnResponse deleteColumn(RpcController controller, DeleteColumnRequest req)
   throws ServiceException {
-    final byte [] tableName = req.getTableName().toByteArray();
-    final byte [] columnName = req.getColumnName().toByteArray();
     try {
-      checkInitialized();
-      if (cpHost != null) {
-        if (cpHost.preDeleteColumn(tableName, columnName)) {
-          return DeleteColumnResponse.newBuilder().build();
-        }
-      }
-      new TableDeleteFamilyHandler(tableName, columnName, this, this).process();
-      if (cpHost != null) {
-        cpHost.postDeleteColumn(tableName, columnName);
-      }
+      deleteColumn(req.getTableName().toByteArray(), req.getColumnName().toByteArray());
     } catch (IOException ioe) {
       throw new ServiceException(ioe);
     }
@@ -1694,20 +1703,23 @@ Server {
   }
 
   @Override
+  public void enableTable(final byte[] tableName) throws IOException {
+    checkInitialized();
+    if (cpHost != null) {
+      cpHost.preEnableTable(tableName);
+    }
+    this.executorService.submit(new EnableTableHandler(this, tableName,
+      catalogTracker, assignmentManager, false));
+    if (cpHost != null) {
+      cpHost.postEnableTable(tableName);
+   }
+  }
+
+  @Override
   public EnableTableResponse enableTable(RpcController controller, EnableTableRequest request)
   throws ServiceException {
-    byte [] tableName = request.getTableName().toByteArray();
     try {
-      checkInitialized();
-      if (cpHost != null) {
-        cpHost.preEnableTable(tableName);
-      }
-      this.executorService.submit(new EnableTableHandler(this, tableName,
-        catalogTracker, assignmentManager, false));
-
-      if (cpHost != null) {
-        cpHost.postEnableTable(tableName);
-     }
+      enableTable(request.getTableName().toByteArray());
     } catch (IOException ioe) {
       throw new ServiceException(ioe);
     }
@@ -1715,20 +1727,23 @@ Server {
   }
 
   @Override
+  public void disableTable(final byte[] tableName) throws IOException {
+    checkInitialized();
+    if (cpHost != null) {
+      cpHost.preDisableTable(tableName);
+    }
+    this.executorService.submit(new DisableTableHandler(this, tableName,
+      catalogTracker, assignmentManager, false));
+    if (cpHost != null) {
+      cpHost.postDisableTable(tableName);
+    }
+  }
+
+  @Override
   public DisableTableResponse disableTable(RpcController controller, DisableTableRequest request)
   throws ServiceException {
-    byte [] tableName = request.getTableName().toByteArray();
     try {
-      checkInitialized();
-      if (cpHost != null) {
-        cpHost.preDisableTable(tableName);
-      }
-      this.executorService.submit(new DisableTableHandler(this, tableName,
-        catalogTracker, assignmentManager, false));
-
-      if (cpHost != null) {
-        cpHost.postDisableTable(tableName);
-      }
+      disableTable(request.getTableName().toByteArray());
     } catch (IOException ioe) {
       throw new ServiceException(ioe);
     }
@@ -1771,25 +1786,28 @@ Server {
   }
 
   @Override
+  public void modifyTable(final byte[] tableName, final HTableDescriptor descriptor)
+      throws IOException {
+    checkInitialized();
+    checkCompression(descriptor);
+    if (cpHost != null) {
+      cpHost.preModifyTable(tableName, descriptor);
+    }
+    new ModifyTableHandler(tableName, descriptor, this, this).process();
+
+    if (cpHost != null) {
+      cpHost.postModifyTable(tableName, descriptor);
+    }
+  }
+
+  @Override
   public ModifyTableResponse modifyTable(RpcController controller, ModifyTableRequest req)
   throws ServiceException {
-    final byte [] tableName = req.getTableName().toByteArray();
-    HTableDescriptor htd = HTableDescriptor.convert(req.getTableSchema());
     try {
-      checkInitialized();
-      checkCompression(htd);
-      if (cpHost != null) {
-        cpHost.preModifyTable(tableName, htd);
-      }
-      TableEventHandler tblHandle = new ModifyTableHandler(tableName, htd, this, this);
-      this.executorService.submit(tblHandle);
-      tblHandle.waitForPersist();
-
-      if (cpHost != null) {
-        cpHost.postModifyTable(tableName, htd);
-      }
+      modifyTable(req.getTableName().toByteArray(),
+        HTableDescriptor.convert(req.getTableSchema()));
     } catch (IOException ioe) {
-        throw new ServiceException(ioe);
+      throw new ServiceException(ioe);
     }
     return ModifyTableResponse.newBuilder().build();
   }
@@ -2344,7 +2362,7 @@ Server {
       String serviceName = call.getServiceName();
       String methodName = call.getMethodName();
       if (!coprocessorServiceHandlers.containsKey(serviceName)) {
-        throw new HBaseRPC.UnknownProtocolException(null,
+        throw new UnknownProtocolException(null,
             "No registered master coprocessor service found for name "+serviceName);
       }
 
@@ -2352,7 +2370,7 @@ Server {
       Descriptors.ServiceDescriptor serviceDesc = service.getDescriptorForType();
       Descriptors.MethodDescriptor methodDesc = serviceDesc.findMethodByName(methodName);
       if (methodDesc == null) {
-        throw new HBaseRPC.UnknownProtocolException(service.getClass(),
+        throw new UnknownProtocolException(service.getClass(),
             "Unknown method "+methodName+" called on master service "+serviceName);
       }
 
@@ -2361,7 +2379,7 @@ Server {
           .mergeFrom(call.getRequest()).build();
       final Message.Builder responseBuilder =
           service.getResponsePrototype(methodDesc).newBuilderForType();
-      service.callMethod(methodDesc, controller, execRequest, new RpcCallback<Message>() {
+      service.callMethod(methodDesc, execController, execRequest, new RpcCallback<Message>() {
         @Override
         public void run(Message message) {
           if (message != null) {
@@ -2582,4 +2600,9 @@ Server {
       throw new ServiceException(e);
     }
   }
+
+  private boolean isHealthCheckerConfigured() {
+    String healthScriptLocation = this.conf.get(HConstants.HEALTH_SCRIPT_LOC);
+    return org.apache.commons.lang.StringUtils.isNotBlank(healthScriptLocation);
+  }
 }

Modified: hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/MasterCoprocessorHost.java
URL: http://svn.apache.org/viewvc/hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/MasterCoprocessorHost.java?rev=1446147&r1=1446146&r2=1446147&view=diff
==============================================================================
--- hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/MasterCoprocessorHost.java (original)
+++ hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/MasterCoprocessorHost.java Thu Feb 14 12:58:12 2013
@@ -64,6 +64,7 @@ public class MasterCoprocessorHost
   private MasterServices masterServices;
 
   MasterCoprocessorHost(final MasterServices services, final Configuration conf) {
+    this.conf = conf;
     this.masterServices = services;
     loadSystemCoprocessors(conf, MASTER_COPROCESSOR_CONF_KEY);
   }

Modified: hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/MasterFileSystem.java
URL: http://svn.apache.org/viewvc/hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/MasterFileSystem.java?rev=1446147&r1=1446146&r2=1446147&view=diff
==============================================================================
--- hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/MasterFileSystem.java (original)
+++ hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/MasterFileSystem.java Thu Feb 14 12:58:12 2013
@@ -32,6 +32,7 @@ import org.apache.hadoop.conf.Configurat
 import org.apache.hadoop.fs.FileStatus;
 import org.apache.hadoop.fs.FileSystem;
 import org.apache.hadoop.fs.Path;
+import org.apache.hadoop.fs.PathFilter;
 import org.apache.hadoop.hbase.ClusterId;
 import org.apache.hadoop.hbase.DeserializationException;
 import org.apache.hadoop.hbase.HColumnDescriptor;
@@ -86,6 +87,18 @@ public class MasterFileSystem {
   final SplitLogManager splitLogManager;
   private final MasterServices services;
 
+  private final static PathFilter META_FILTER = new PathFilter() {
+    public boolean accept(Path p) {
+      return HLogUtil.isMetaFile(p);
+    }
+  };
+
+  private final static PathFilter NON_META_FILTER = new PathFilter() {
+    public boolean accept(Path p) {
+      return !HLogUtil.isMetaFile(p);
+    }
+  };
+
   public MasterFileSystem(Server master, MasterServices services,
       MetricsMaster metricsMaster, boolean masterRecovery)
   throws IOException {
@@ -242,7 +255,8 @@ public class MasterFileSystem {
                 + " belongs to an existing region server");
           }
         }
-        splitLog(serverNames);
+        splitLog(serverNames, META_FILTER);
+        splitLog(serverNames, NON_META_FILTER);
         retrySplitting = false;
       } catch (IOException ioe) {
         LOG.warn("Failed splitting of " + serverNames, ioe);
@@ -271,8 +285,30 @@ public class MasterFileSystem {
     splitLog(serverNames);
   }
 
-  public void splitLog(final List<ServerName> serverNames) throws IOException {
+  /**
+   * Specialized method to handle the splitting for meta HLog
+   * @param serverName
+   * @throws IOException
+   */
+  public void splitMetaLog(final ServerName serverName) throws IOException {
     long splitTime = 0, splitLogSize = 0;
+    List<ServerName> serverNames = new ArrayList<ServerName>();
+    serverNames.add(serverName);
+    List<Path> logDirs = getLogDirs(serverNames);
+    if (logDirs.isEmpty()) {
+      LOG.info("No meta logs to split");
+      return;
+    }
+    splitLogManager.handleDeadWorkers(serverNames);
+    splitTime = EnvironmentEdgeManager.currentTimeMillis();
+    splitLogSize = splitLogManager.splitLogDistributed(logDirs, META_FILTER);
+    splitTime = EnvironmentEdgeManager.currentTimeMillis() - splitTime;
+    if (this.metricsMaster != null) {
+      this.metricsMaster.addSplit(splitTime, splitLogSize);
+    }
+  }
+
+  private List<Path> getLogDirs(final List<ServerName> serverNames) throws IOException {
     List<Path> logDirs = new ArrayList<Path>();
     for (ServerName serverName: serverNames) {
       Path logDir = new Path(this.rootdir, HLogUtil.getHLogDirectoryName(serverName.toString()));
@@ -290,6 +326,23 @@ public class MasterFileSystem {
       }
       logDirs.add(splitDir);
     }
+    return logDirs;
+  }
+
+  public void splitLog(final List<ServerName> serverNames) throws IOException {
+    splitLog(serverNames, NON_META_FILTER);
+  }
+
+  /**
+   * This method is the base split method that splits HLog files matching a filter.
+   * Callers should pass the appropriate filter for meta and non-meta HLogs.
+   * @param serverNames
+   * @param filter
+   * @throws IOException
+   */
+  public void splitLog(final List<ServerName> serverNames, PathFilter filter) throws IOException {
+    long splitTime = 0, splitLogSize = 0;
+    List<Path> logDirs = getLogDirs(serverNames);
 
     if (logDirs.isEmpty()) {
       LOG.info("No logs to split");
@@ -299,7 +352,7 @@ public class MasterFileSystem {
     if (distributedLogSplitting) {
       splitLogManager.handleDeadWorkers(serverNames);
       splitTime = EnvironmentEdgeManager.currentTimeMillis();
-      splitLogSize = splitLogManager.splitLogDistributed(logDirs);
+      splitLogSize = splitLogManager.splitLogDistributed(logDirs,filter);
       splitTime = EnvironmentEdgeManager.currentTimeMillis() - splitTime;
     } else {
       for(Path logDir: logDirs){
@@ -411,7 +464,7 @@ public class MasterFileSystem {
       // if not the cleaner will take care of them.
       for (Path tabledir: FSUtils.getTableDirs(fs, tmpdir)) {
         for (Path regiondir: FSUtils.getRegionDirs(fs, tabledir)) {
-          HFileArchiver.archiveRegion(fs, this.rootdir, tabledir, regiondir);
+          HFileArchiver.archiveRegion(c, fs, this.rootdir, tabledir, regiondir);
         }
       }
       if (!fs.delete(tmpdir, true)) {

Modified: hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/MasterServices.java
URL: http://svn.apache.org/viewvc/hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/MasterServices.java?rev=1446147&r1=1446146&r2=1446147&view=diff
==============================================================================
--- hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/MasterServices.java (original)
+++ hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/MasterServices.java Thu Feb 14 12:58:12 2013
@@ -22,6 +22,7 @@ import java.io.IOException;
 
 import com.google.protobuf.Service;
 import org.apache.hadoop.classification.InterfaceAudience;
+import org.apache.hadoop.hbase.HColumnDescriptor;
 import org.apache.hadoop.hbase.HTableDescriptor;
 import org.apache.hadoop.hbase.Server;
 import org.apache.hadoop.hbase.TableDescriptors;
@@ -80,6 +81,63 @@ public interface MasterServices extends 
       throws IOException;
 
   /**
+   * Delete a table
+   * @param tableName The table name
+   * @throws IOException
+   */
+  public void deleteTable(final byte[] tableName) throws IOException;
+
+  /**
+   * Modify the descriptor of an existing table
+   * @param tableName The table name
+   * @param descriptor The updated table descriptor
+   * @throws IOException
+   */
+  public void modifyTable(final byte[] tableName, final HTableDescriptor descriptor)
+      throws IOException;
+
+  /**
+   * Enable an existing table
+   * @param tableName The table name
+   * @throws IOException
+   */
+  public void enableTable(final byte[] tableName) throws IOException;
+
+  /**
+   * Disable an existing table
+   * @param tableName The table name
+   * @throws IOException
+   */
+  public void disableTable(final byte[] tableName) throws IOException;
+
+  /**
+   * Add a new column to an existing table
+   * @param tableName The table name
+   * @param column The column definition
+   * @throws IOException
+   */
+  public void addColumn(final byte[] tableName, final HColumnDescriptor column)
+      throws IOException;
+
+  /**
+   * Modify the column descriptor of an existing column in an existing table
+   * @param tableName The table name
+   * @param descriptor The updated column definition
+   * @throws IOException
+   */
+  public void modifyColumn(byte[] tableName, HColumnDescriptor descriptor)
+      throws IOException;
+
+  /**
+   * Delete a column from an existing table
+   * @param tableName The table name
+   * @param columnName The column name
+   * @throws IOException
+   */
+  public void deleteColumn(final byte[] tableName, final byte[] columnName)
+      throws IOException;
+
+  /**
    * @return Return table descriptors implementation.
    */
   public TableDescriptors getTableDescriptors();

Modified: hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/RegionPlan.java
URL: http://svn.apache.org/viewvc/hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/RegionPlan.java?rev=1446147&r1=1446146&r2=1446147&view=diff
==============================================================================
--- hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/RegionPlan.java (original)
+++ hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/RegionPlan.java Thu Feb 14 12:58:12 2013
@@ -117,6 +117,18 @@ public class RegionPlan implements Compa
   }
 
   @Override
+  public boolean equals(Object obj) {
+    if (this == obj) {
+      return true;
+    }
+    if (obj == null || getClass() != obj.getClass()) {
+      return false;
+    }
+    RegionPlan other = (RegionPlan) obj;
+    return compareTo(other) == 0;
+  }
+
+  @Override
   public String toString() {
     return "hri=" + this.hri.getRegionNameAsString() + ", src=" +
       (this.source == null? "": this.source.toString()) +

Modified: hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/ServerManager.java
URL: http://svn.apache.org/viewvc/hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/ServerManager.java?rev=1446147&r1=1446146&r2=1446147&view=diff
==============================================================================
--- hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/ServerManager.java (original)
+++ hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/ServerManager.java Thu Feb 14 12:58:12 2013
@@ -380,7 +380,7 @@ public class ServerManager {
   public double getAverageLoad() {
     int totalLoad = 0;
     int numServers = 0;
-    double averageLoad = 0.0;
+    double averageLoad;
     for (ServerLoad sl: this.onlineServers.values()) {
         numServers++;
         totalLoad += sl.getNumberOfRegions();
@@ -680,7 +680,7 @@ public class ServerManager {
     */
   private AdminProtocol getServerConnection(final ServerName sn)
   throws IOException {
-    AdminProtocol admin = this.serverConnections.get(sn.toString());
+    AdminProtocol admin = this.serverConnections.get(sn);
     if (admin == null) {
       LOG.debug("New connection to " + sn.toString());
       admin = this.connection.getAdmin(sn.getHostname(), sn.getPort());
@@ -886,7 +886,7 @@ public class ServerManager {
    * To clear any dead server with same host name and port of any online server
    */
   void clearDeadServersWithSameHostNameAndPortOfOnlineServer() {
-    ServerName sn = null;
+    ServerName sn;
     for (ServerName serverName : getOnlineServersList()) {
       while ((sn = ServerName.
           findServerWithSameHostnamePort(this.deadservers, serverName)) != null) {

Modified: hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/SplitLogManager.java
URL: http://svn.apache.org/viewvc/hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/SplitLogManager.java?rev=1446147&r1=1446146&r2=1446147&view=diff
==============================================================================
--- hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/SplitLogManager.java (original)
+++ hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/SplitLogManager.java Thu Feb 14 12:58:12 2013
@@ -26,6 +26,7 @@ import static org.apache.hadoop.hbase.ma
 
 import java.io.IOException;
 import java.util.ArrayList;
+import java.util.Collections;
 import java.util.HashSet;
 import java.util.List;
 import java.util.Map;
@@ -40,10 +41,11 @@ import org.apache.hadoop.conf.Configurat
 import org.apache.hadoop.fs.FileStatus;
 import org.apache.hadoop.fs.FileSystem;
 import org.apache.hadoop.fs.Path;
+import org.apache.hadoop.fs.PathFilter;
 import org.apache.hadoop.hbase.Chore;
-import org.apache.hadoop.hbase.SplitLogCounters;
 import org.apache.hadoop.hbase.DeserializationException;
 import org.apache.hadoop.hbase.ServerName;
+import org.apache.hadoop.hbase.SplitLogCounters;
 import org.apache.hadoop.hbase.SplitLogTask;
 import org.apache.hadoop.hbase.Stoppable;
 import org.apache.hadoop.hbase.master.SplitLogManager.TaskFinisher.Status;
@@ -123,6 +125,8 @@ public class SplitLogManager extends Zoo
   private volatile Set<ServerName> deadWorkers = null;
   private final Object deadWorkersLock = new Object();
 
+  private Set<String> failedDeletions = null;
+
   /**
    * Wrapper around {@link #SplitLogManager(ZooKeeperWatcher zkw, Configuration conf,
    *   Stoppable stopper, MasterServices master, ServerName serverName, TaskFinisher tf)}
@@ -180,6 +184,8 @@ public class SplitLogManager extends Zoo
     this.serverName = serverName;
     this.timeoutMonitor =
       new TimeoutMonitor(conf.getInt("hbase.splitlog.manager.timeoutmonitor.period", 1000), stopper);
+
+    this.failedDeletions = Collections.synchronizedSet(new HashSet<String>());
   }
 
   public void finishInitialization(boolean masterRecovery) {
@@ -194,7 +200,7 @@ public class SplitLogManager extends Zoo
     }
   }
 
-  private FileStatus[] getFileList(List<Path> logDirs) throws IOException {
+  private FileStatus[] getFileList(List<Path> logDirs, PathFilter filter) throws IOException {
     List<FileStatus> fileStatus = new ArrayList<FileStatus>();
     for (Path hLogDir : logDirs) {
       this.fs = hLogDir.getFileSystem(conf);
@@ -202,8 +208,7 @@ public class SplitLogManager extends Zoo
         LOG.warn(hLogDir + " doesn't exist. Nothing to do!");
         continue;
       }
-      // TODO filter filenames?
-      FileStatus[] logfiles = FSUtils.listStatus(fs, hLogDir, null);
+      FileStatus[] logfiles = FSUtils.listStatus(fs, hLogDir, filter);
       if (logfiles == null || logfiles.length == 0) {
         LOG.info(hLogDir + " is empty dir, no logs to split");
       } else {
@@ -228,6 +233,7 @@ public class SplitLogManager extends Zoo
     logDirs.add(logDir);
     return splitLogDistributed(logDirs);
   }
+
   /**
    * The caller will block until all the log files of the given region server
    * have been processed - successfully split or an error is encountered - by an
@@ -239,9 +245,25 @@ public class SplitLogManager extends Zoo
    * @return cumulative size of the logfiles split
    */
   public long splitLogDistributed(final List<Path> logDirs) throws IOException {
+    return splitLogDistributed(logDirs, null);
+  }
+
+  /**
+   * The caller will block until all the META log files of the given region server
+   * have been processed - successfully split or an error is encountered - by an
+   * available worker region server. This method must only be called after the
+   * region servers have been brought online.
+   *
+   * @param logDirs List of log dirs to split
+   * @param filter the Path filter to select specific files for considering
+   * @throws IOException If there was an error while splitting any log file
+   * @return cumulative size of the logfiles split
+   */
+  public long splitLogDistributed(final List<Path> logDirs, PathFilter filter) 
+      throws IOException {
     MonitoredTask status = TaskMonitor.get().createStatus(
           "Doing distributed log split in " + logDirs);
-    FileStatus[] logfiles = getFileList(logDirs);
+    FileStatus[] logfiles = getFileList(logDirs, filter);
     status.setStatus("Checking directory contents...");
     LOG.debug("Scheduling batch of logs to split");
     SplitLogCounters.tot_mgr_log_split_batch_start.incrementAndGet();
@@ -418,11 +440,12 @@ public class SplitLogManager extends Zoo
         }
       }
     }
-    // delete the task node in zk. Keep trying indefinitely - its an async
+    // delete the task node in zk. It's an async
     // call and no one is blocked waiting for this node to be deleted. All
     // task names are unique (log.<timestamp>) there is no risk of deleting
     // a future task.
-    deleteNode(path, Long.MAX_VALUE);
+    // if a deletion fails, TimeoutMonitor will retry the same deletion later
+    deleteNode(path, zkretries);
     return;
   }
 
@@ -531,6 +554,21 @@ public class SplitLogManager extends Zoo
     }
   }
 
+  /**
+   * Helper function to check whether to abandon retries in ZooKeeper AsyncCallback functions
+   * @param statusCode integer value of a ZooKeeper exception code
+   * @param action description message about the retried action
+   * @return true when need to abandon retries otherwise false
+   */
+  private boolean needAbandonRetries(int statusCode, String action) {
+    if (statusCode == KeeperException.Code.SESSIONEXPIRED.intValue()) {
+      LOG.error("ZK session expired. Master is expected to shut down. Abandoning retries for "
+          + "action=" + action);
+      return true;
+    }
+    return false;
+  }
+
   private void heartbeat(String path, int new_version, ServerName workerName) {
     Task task = findOrCreateOrphanTask(path);
     if (new_version != task.last_version) {
@@ -662,8 +700,7 @@ public class SplitLogManager extends Zoo
   }
 
   private void deleteNodeFailure(String path) {
-    LOG.fatal("logic failure, failing to delete a node should never happen " +
-        "because delete has infinite retries");
+    LOG.info("Failed to delete node " + path + " and will retry soon.");
     return;
   }
 
@@ -847,7 +884,7 @@ public class SplitLogManager extends Zoo
     volatile long last_update;
     volatile int last_version;
     volatile ServerName cur_worker_name;
-    TaskBatch batch;
+    volatile TaskBatch batch;
     volatile TerminationStatus status;
     volatile int incarnation;
     volatile int unforcedResubmits;
@@ -1005,6 +1042,16 @@ public class SplitLogManager extends Zoo
         SplitLogCounters.tot_mgr_resubmit_unassigned.incrementAndGet();
         LOG.debug("resubmitting unassigned task(s) after timeout");
       }
+
+      // Retry previously failed deletes
+      if (failedDeletions.size() > 0) {
+        List<String> tmpPaths = new ArrayList<String>(failedDeletions);
+        for (String tmpPath : tmpPaths) {
+          // deleteNode is an async call
+          deleteNode(tmpPath, zkretries);
+        }
+        failedDeletions.removeAll(tmpPaths);
+      }
     }
   }
 
@@ -1019,6 +1066,10 @@ public class SplitLogManager extends Zoo
     public void processResult(int rc, String path, Object ctx, String name) {
       SplitLogCounters.tot_mgr_node_create_result.incrementAndGet();
       if (rc != 0) {
+        if (needAbandonRetries(rc, "Create znode " + path)) {
+          createNodeFailure(path);
+          return;
+        }
         if (rc == KeeperException.Code.NODEEXISTS.intValue()) {
           // What if there is a delete pending against this pre-existing
           // znode? Then this soon-to-be-deleted task znode must be in TASK_DONE
@@ -1058,8 +1109,7 @@ public class SplitLogManager extends Zoo
         Stat stat) {
       SplitLogCounters.tot_mgr_get_data_result.incrementAndGet();
       if (rc != 0) {
-        if (rc == KeeperException.Code.SESSIONEXPIRED.intValue()) {
-          LOG.error("ZK session expired. Master is expected to shut down. Abandoning retries.");
+        if (needAbandonRetries(rc, "GetData from znode " + path)) {
           return;
         }
         if (rc == KeeperException.Code.NONODE.intValue()) {
@@ -1113,6 +1163,10 @@ public class SplitLogManager extends Zoo
     public void processResult(int rc, String path, Object ctx) {
       SplitLogCounters.tot_mgr_node_delete_result.incrementAndGet();
       if (rc != 0) {
+        if (needAbandonRetries(rc, "Delete znode " + path)) {
+          failedDeletions.add(path);
+          return;
+        }
         if (rc != KeeperException.Code.NONODE.intValue()) {
           SplitLogCounters.tot_mgr_node_delete_err.incrementAndGet();
           Long retry_count = (Long) ctx;
@@ -1120,13 +1174,14 @@ public class SplitLogManager extends Zoo
               path + " remaining retries=" + retry_count);
           if (retry_count == 0) {
             LOG.warn("delete failed " + path);
+            failedDeletions.add(path);
             deleteNodeFailure(path);
           } else {
             deleteNode(path, retry_count - 1);
           }
           return;
         } else {
-        LOG.debug(path +
+          LOG.info(path +
             " does not exist. Either was created but deleted behind our" +
             " back by another pending delete OR was deleted" +
             " in earlier retry rounds. zkretries = " + (Long) ctx);
@@ -1151,8 +1206,7 @@ public class SplitLogManager extends Zoo
     @Override
     public void processResult(int rc, String path, Object ctx, String name) {
       if (rc != 0) {
-        if (rc == KeeperException.Code.SESSIONEXPIRED.intValue()) {
-          LOG.error("ZK session expired. Master is expected to shut down. Abandoning retries.");
+        if (needAbandonRetries(rc, "CreateRescan znode " + path)) {
           return;
         }
         Long retry_count = (Long)ctx;

Modified: hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/cleaner/CleanerChore.java
URL: http://svn.apache.org/viewvc/hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/cleaner/CleanerChore.java?rev=1446147&r1=1446146&r2=1446147&view=diff
==============================================================================
--- hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/cleaner/CleanerChore.java (original)
+++ hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/cleaner/CleanerChore.java Thu Feb 14 12:58:12 2013
@@ -145,13 +145,23 @@ public abstract class CleanerChore<T ext
    * @return <tt>true</tt> if the directory was deleted, <tt>false</tt> otherwise.
    * @throws IOException if there is an unexpected filesystem error
    */
-  private boolean checkAndDeleteDirectory(Path toCheck) throws IOException {
+  public boolean checkAndDeleteDirectory(Path toCheck) throws IOException {
     if (LOG.isTraceEnabled()) {
       LOG.trace("Checking directory: " + toCheck);
     }
     FileStatus[] children = FSUtils.listStatus(fs, toCheck);
     // if the directory doesn't exist, then we are done
-    if (children == null) return true;
+    if (children == null) {
+      try {
+        return fs.delete(toCheck, false);
+      } catch (IOException e) {
+        if (LOG.isTraceEnabled()) {
+          LOG.trace("Couldn't delete directory: " + toCheck, e);
+        }
+      }
+      // couldn't delete w/o exception, so we can't return success.
+      return false;
+    }
 
     boolean canDeleteThis = true;
     for (FileStatus child : children) {
@@ -168,9 +178,22 @@ public abstract class CleanerChore<T ext
       }
     }
 
-    // if all the children have been deleted, then we should try to delete this directory. However,
-    // don't do so recursively so we don't delete files that have been added since we checked.
-    return canDeleteThis ? fs.delete(toCheck, false) : false;
+    // if the directory has children, we can't delete it, so we are done
+    if (!canDeleteThis) return false;
+
+    // otherwise, all the children (that we know about) have been deleted, so we should try to
+    // delete this directory. However, don't do so recursively so we don't delete files that have
+    // been added since we last checked.
+    try {
+      return fs.delete(toCheck, false);
+    } catch (IOException e) {
+      if (LOG.isTraceEnabled()) {
+        LOG.trace("Couldn't delete directory: " + toCheck, e);
+      }
+    }
+
+    // couldn't delete w/o exception, so we can't return success.
+    return false;
   }
 
   /**

Modified: hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/cleaner/TimeToLiveHFileCleaner.java
URL: http://svn.apache.org/viewvc/hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/cleaner/TimeToLiveHFileCleaner.java?rev=1446147&r1=1446146&r2=1446147&view=diff
==============================================================================
--- hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/cleaner/TimeToLiveHFileCleaner.java (original)
+++ hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/cleaner/TimeToLiveHFileCleaner.java Thu Feb 14 12:58:12 2013
@@ -30,7 +30,7 @@ import org.apache.hadoop.hbase.util.Envi
 
 /**
  * HFile cleaner that uses the timestamp of the hfile to determine if it should be deleted. By
- * default they are allowed to live for {@value TimeToLiveHFileCleaner#DEFAULT_TTL}
+ * default they are allowed to live for {@value #DEFAULT_TTL}
  */
 @InterfaceAudience.Private
 public class TimeToLiveHFileCleaner extends BaseHFileCleanerDelegate {
@@ -38,7 +38,7 @@ public class TimeToLiveHFileCleaner exte
   public static final Log LOG = LogFactory.getLog(TimeToLiveHFileCleaner.class.getName());
   public static final String TTL_CONF_KEY = "hbase.master.hfilecleaner.ttl";
   // default ttl = 5 minutes
-  private static final long DEFAULT_TTL = 60000 * 5;
+  public static final long DEFAULT_TTL = 60000 * 5;
   // Configured time a hfile can be kept after it was moved to the archive
   private long ttl;
   private FileSystem fs;

Modified: hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/handler/DeleteTableHandler.java
URL: http://svn.apache.org/viewvc/hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/handler/DeleteTableHandler.java?rev=1446147&r1=1446146&r2=1446147&view=diff
==============================================================================
--- hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/handler/DeleteTableHandler.java (original)
+++ hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/handler/DeleteTableHandler.java Thu Feb 14 12:58:12 2013
@@ -97,7 +97,7 @@ public class DeleteTableHandler extends 
     FileSystem fs = mfs.getFileSystem();
     for (HRegionInfo hri: regions) {
       LOG.debug("Deleting region " + hri.getRegionNameAsString() + " from FS");
-      HFileArchiver.archiveRegion(fs, mfs.getRootDir(),
+      HFileArchiver.archiveRegion(masterServices.getConfiguration(), fs, mfs.getRootDir(),
           tempTableDir, new Path(tempTableDir, hri.getEncodedName()));
     }
 

Modified: hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/handler/MetaServerShutdownHandler.java
URL: http://svn.apache.org/viewvc/hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/handler/MetaServerShutdownHandler.java?rev=1446147&r1=1446146&r2=1446147&view=diff
==============================================================================
--- hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/handler/MetaServerShutdownHandler.java (original)
+++ hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/handler/MetaServerShutdownHandler.java Thu Feb 14 12:58:12 2013
@@ -18,11 +18,17 @@
  */
 package org.apache.hadoop.hbase.master.handler;
 
+import java.io.IOException;
+
+import org.apache.commons.logging.Log;
+import org.apache.commons.logging.LogFactory;
 import org.apache.hadoop.classification.InterfaceAudience;
+import org.apache.hadoop.hbase.HRegionInfo;
 import org.apache.hadoop.hbase.Server;
 import org.apache.hadoop.hbase.ServerName;
 import org.apache.hadoop.hbase.master.DeadServer;
 import org.apache.hadoop.hbase.master.MasterServices;
+import org.apache.zookeeper.KeeperException;
 
 /**
  * Shutdown handler for the server hosting <code>-ROOT-</code>,
@@ -32,7 +38,7 @@ import org.apache.hadoop.hbase.master.Ma
 public class MetaServerShutdownHandler extends ServerShutdownHandler {
   private final boolean carryingRoot;
   private final boolean carryingMeta;
-
+  private static final Log LOG = LogFactory.getLog(MetaServerShutdownHandler.class);
   public MetaServerShutdownHandler(final Server server,
       final MasterServices services,
       final DeadServer deadServers, final ServerName serverName,
@@ -44,11 +50,118 @@ public class MetaServerShutdownHandler e
   }
 
   @Override
+  public void process() throws IOException {
+    try {
+      LOG.info("Splitting META logs for " + serverName);
+      if (this.shouldSplitHlog) {
+        this.services.getMasterFileSystem().splitMetaLog(serverName);
+      }
+    } catch (IOException ioe) {
+      this.services.getExecutorService().submit(this);
+      this.deadServers.add(serverName);
+      throw new IOException("failed log splitting for " +
+          serverName + ", will retry", ioe);
+    }
+
+    // Assign root and meta if we were carrying them.
+    if (isCarryingRoot()) { // -ROOT-
+      // Check again: region may be assigned to other where because of RIT
+      // timeout
+      if (this.services.getAssignmentManager().isCarryingRoot(serverName)) {
+        LOG.info("Server " + serverName
+            + " was carrying ROOT. Trying to assign.");
+        this.services.getAssignmentManager().regionOffline(
+            HRegionInfo.ROOT_REGIONINFO);
+        verifyAndAssignRootWithRetries();
+      } else {
+        LOG.info("ROOT has been assigned to otherwhere, skip assigning.");
+      }
+    }
+
+    // Carrying meta?
+    if (isCarryingMeta()) {
+      // Check again: region may be assigned to other where because of RIT
+      // timeout
+      if (this.services.getAssignmentManager().isCarryingMeta(serverName)) {
+        LOG.info("Server " + serverName
+            + " was carrying META. Trying to assign.");
+        this.services.getAssignmentManager().regionOffline(
+            HRegionInfo.FIRST_META_REGIONINFO);
+        this.services.getAssignmentManager().assignMeta();
+      } else {
+        LOG.info("META has been assigned to otherwhere, skip assigning.");
+      }
+
+    }
+    super.process();
+  }
+  /**
+   * Before assign the ROOT region, ensure it haven't
+   *  been assigned by other place
+   * <p>
+   * Under some scenarios, the ROOT region can be opened twice, so it seemed online
+   * in two regionserver at the same time.
+   * If the ROOT region has been assigned, so the operation can be canceled.
+   * @throws InterruptedException
+   * @throws IOException
+   * @throws KeeperException
+   */
+  private void verifyAndAssignRoot()
+  throws InterruptedException, IOException, KeeperException {
+    long timeout = this.server.getConfiguration().
+      getLong("hbase.catalog.verification.timeout", 1000);
+    if (!this.server.getCatalogTracker().verifyRootRegionLocation(timeout)) {
+      this.services.getAssignmentManager().assignRoot();
+    } else if (serverName.equals(server.getCatalogTracker().getRootLocation())) {
+      throw new IOException("-ROOT- is onlined on the dead server "
+          + serverName);
+    } else {
+      LOG.info("Skip assigning -ROOT-, because it is online on the "
+          + server.getCatalogTracker().getRootLocation());
+    }
+  }
+
+  /**
+   * Failed many times, shutdown processing
+   * @throws IOException
+   */
+  private void verifyAndAssignRootWithRetries() throws IOException {
+    int iTimes = this.server.getConfiguration().getInt(
+        "hbase.catalog.verification.retries", 10);
+
+    long waitTime = this.server.getConfiguration().getLong(
+        "hbase.catalog.verification.timeout", 1000);
+
+    int iFlag = 0;
+    while (true) {
+      try {
+        verifyAndAssignRoot();
+        break;
+      } catch (KeeperException e) {
+        this.server.abort("In server shutdown processing, assigning root", e);
+        throw new IOException("Aborting", e);
+      } catch (Exception e) {
+        if (iFlag >= iTimes) {
+          this.server.abort("verifyAndAssignRoot failed after" + iTimes
+              + " times retries, aborting", e);
+          throw new IOException("Aborting", e);
+        }
+        try {
+          Thread.sleep(waitTime);
+        } catch (InterruptedException e1) {
+          LOG.warn("Interrupted when is the thread sleep", e1);
+          Thread.currentThread().interrupt();
+          throw new IOException("Interrupted", e1);
+        }
+        iFlag++;
+      }
+    }
+  }
+
   boolean isCarryingRoot() {
     return this.carryingRoot;
   }
 
-  @Override
   boolean isCarryingMeta() {
     return this.carryingMeta;
   }

Modified: hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/handler/ServerShutdownHandler.java
URL: http://svn.apache.org/viewvc/hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/handler/ServerShutdownHandler.java?rev=1446147&r1=1446146&r2=1446147&view=diff
==============================================================================
--- hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/handler/ServerShutdownHandler.java (original)
+++ hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/handler/ServerShutdownHandler.java Thu Feb 14 12:58:12 2013
@@ -55,10 +55,10 @@ import org.apache.zookeeper.KeeperExcept
 @InterfaceAudience.Private
 public class ServerShutdownHandler extends EventHandler {
   private static final Log LOG = LogFactory.getLog(ServerShutdownHandler.class);
-  private final ServerName serverName;
-  private final MasterServices services;
-  private final DeadServer deadServers;
-  private final boolean shouldSplitHlog; // whether to split HLog or not
+  protected final ServerName serverName;
+  protected final MasterServices services;
+  protected final DeadServer deadServers;
+  protected final boolean shouldSplitHlog; // whether to split HLog or not
 
   public ServerShutdownHandler(final Server server, final MasterServices services,
       final DeadServer deadServers, final ServerName serverName,
@@ -91,63 +91,6 @@ public class ServerShutdownHandler exten
   }
 
   /**
-   * Before assign the ROOT region, ensure it haven't
-   *  been assigned by other place
-   * <p>
-   * Under some scenarios, the ROOT region can be opened twice, so it seemed online
-   * in two regionserver at the same time.
-   * If the ROOT region has been assigned, so the operation can be canceled.
-   * @throws InterruptedException
-   * @throws IOException
-   * @throws KeeperException
-   */
-  private void verifyAndAssignRoot()
-  throws InterruptedException, IOException, KeeperException {
-    long timeout = this.server.getConfiguration().
-      getLong("hbase.catalog.verification.timeout", 1000);
-    if (!this.server.getCatalogTracker().verifyRootRegionLocation(timeout)) {
-      this.services.getAssignmentManager().assignRoot();
-    }
-  }
-
-  /**
-   * Failed many times, shutdown processing
-   * @throws IOException
-   */
-  private void verifyAndAssignRootWithRetries() throws IOException {
-    int iTimes = this.server.getConfiguration().getInt(
-        "hbase.catalog.verification.retries", 10);
-
-    long waitTime = this.server.getConfiguration().getLong(
-        "hbase.catalog.verification.timeout", 1000);
-
-    int iFlag = 0;
-    while (true) {
-      try {
-        verifyAndAssignRoot();
-        break;
-      } catch (KeeperException e) {
-        this.server.abort("In server shutdown processing, assigning root", e);
-        throw new IOException("Aborting", e);
-      } catch (Exception e) {
-        if (iFlag >= iTimes) {
-          this.server.abort("verifyAndAssignRoot failed after" + iTimes
-              + " times retries, aborting", e);
-          throw new IOException("Aborting", e);
-        }
-        try {
-          Thread.sleep(waitTime);
-        } catch (InterruptedException e1) {
-          LOG.warn("Interrupted when is the thread sleep", e1);
-          Thread.currentThread().interrupt();
-          throw new IOException("Interrupted", e1);
-        }
-        iFlag++;
-      }
-    }
-  }
-
-  /**
    * @return True if the server we are processing was carrying <code>-ROOT-</code>
    */
   boolean isCarryingRoot() {
@@ -182,30 +125,13 @@ public class ServerShutdownHandler exten
           LOG.info("Skipping log splitting for " + serverName);
         }
       } catch (IOException ioe) {
-        this.services.getExecutorService().submit(this);
+        //typecast to SSH so that we make sure that it is the SSH instance that
+        //gets submitted as opposed to MSSH or some other derived instance of SSH
+        this.services.getExecutorService().submit((ServerShutdownHandler)this);
         this.deadServers.add(serverName);
         throw new IOException("failed log splitting for " +
           serverName + ", will retry", ioe);
       }
-
-      // Assign root and meta if we were carrying them.
-      if (isCarryingRoot()) { // -ROOT-
-        LOG.info("Server " + serverName +
-            " was carrying ROOT. Trying to assign.");
-        this.services.getAssignmentManager().
-          regionOffline(HRegionInfo.ROOT_REGIONINFO);
-        verifyAndAssignRootWithRetries();
-      }
-
-      // Carrying meta?
-      if (isCarryingMeta()) {
-        LOG.info("Server " + serverName +
-          " was carrying META. Trying to assign.");
-        this.services.getAssignmentManager().
-          regionOffline(HRegionInfo.FIRST_META_REGIONINFO);
-        this.services.getAssignmentManager().assignMeta();
-      }
-
       // We don't want worker thread in the MetaServerShutdownHandler
       // executor pool to block by waiting availability of -ROOT-
       // and .META. server. Otherwise, it could run into the following issue:
@@ -430,7 +356,7 @@ public class ServerShutdownHandler exten
     if (daughter == null) return 0;
     if (isDaughterMissing(catalogTracker, daughter)) {
       LOG.info("Fixup; missing daughter " + daughter.getRegionNameAsString());
-      MetaEditor.addDaughter(catalogTracker, daughter, null);
+      MetaEditor.addDaughter(catalogTracker, daughter, null, HConstants.NO_SEQNUM);
 
       // TODO: Log WARN if the regiondir does not exist in the fs.  If its not
       // there then something wonky about the split -- things will keep going

Modified: hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/handler/SplitRegionHandler.java
URL: http://svn.apache.org/viewvc/hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/handler/SplitRegionHandler.java?rev=1446147&r1=1446146&r2=1446147&view=diff
==============================================================================
--- hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/handler/SplitRegionHandler.java (original)
+++ hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/handler/SplitRegionHandler.java Thu Feb 14 12:58:12 2013
@@ -46,6 +46,7 @@ public class SplitRegionHandler extends 
   /**
    * For testing only!  Set to true to skip handling of split.
    */
+  @edu.umd.cs.findbugs.annotations.SuppressWarnings(value="MS_SHOULD_BE_FINAL")
   public static boolean TEST_SKIP = false;
 
   public SplitRegionHandler(Server server,

Modified: hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/handler/TableAddFamilyHandler.java
URL: http://svn.apache.org/viewvc/hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/handler/TableAddFamilyHandler.java?rev=1446147&r1=1446146&r2=1446147&view=diff
==============================================================================
--- hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/handler/TableAddFamilyHandler.java (original)
+++ hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/handler/TableAddFamilyHandler.java Thu Feb 14 12:58:12 2013
@@ -58,15 +58,13 @@ public class TableAddFamilyHandler exten
     if(cpHost != null){
       cpHost.preAddColumnHandler(this.tableName, this.familyDesc);
     }
-    // Update table descriptor in HDFS
-    HTableDescriptor htd = this.masterServices.getMasterFileSystem()
-        .addColumn(tableName, familyDesc);
-    // Update in-memory descriptor cache
-    this.masterServices.getTableDescriptors().add(htd);
+    // Update table descriptor
+    this.masterServices.getMasterFileSystem().addColumn(tableName, familyDesc);
     if(cpHost != null){
       cpHost.postAddColumnHandler(this.tableName, this.familyDesc);
     }
   }
+
   @Override
   public String toString() {
     String name = "UnknownServerName";

Modified: hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/handler/TableDeleteFamilyHandler.java
URL: http://svn.apache.org/viewvc/hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/handler/TableDeleteFamilyHandler.java?rev=1446147&r1=1446146&r2=1446147&view=diff
==============================================================================
--- hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/handler/TableDeleteFamilyHandler.java (original)
+++ hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/handler/TableDeleteFamilyHandler.java Thu Feb 14 12:58:12 2013
@@ -53,11 +53,8 @@ public class TableDeleteFamilyHandler ex
     if (cpHost != null) {
       cpHost.preDeleteColumnHandler(this.tableName, this.familyName);
     }
-    // Update table descriptor in HDFS
-    HTableDescriptor htd =
-      this.masterServices.getMasterFileSystem().deleteColumn(tableName, familyName);
-    // Update in-memory descriptor cache
-    this.masterServices.getTableDescriptors().add(htd);
+    // Update table descriptor
+    this.masterServices.getMasterFileSystem().deleteColumn(tableName, familyName);
     // Remove the column family from the file system
     MasterFileSystem mfs = this.masterServices.getMasterFileSystem();
     for (HRegionInfo hri : hris) {

Modified: hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/handler/TableEventHandler.java
URL: http://svn.apache.org/viewvc/hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/handler/TableEventHandler.java?rev=1446147&r1=1446146&r2=1446147&view=diff
==============================================================================
--- hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/handler/TableEventHandler.java (original)
+++ hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/handler/TableEventHandler.java Thu Feb 14 12:58:12 2013
@@ -61,7 +61,6 @@ public abstract class TableEventHandler 
   protected final MasterServices masterServices;
   protected final byte [] tableName;
   protected final String tableNameStr;
-  protected boolean persistedToZk = false;
 
   public TableEventHandler(EventType eventType, byte [] tableName, Server server,
       MasterServices masterServices)
@@ -111,10 +110,7 @@ public abstract class TableEventHandler 
       LOG.error("Error manipulating table " + Bytes.toString(tableName), e);
     } catch (KeeperException e) {
       LOG.error("Error manipulating table " + Bytes.toString(tableName), e);
-    } finally {
-      // notify the waiting thread that we're done persisting the request
-      setPersist();
-    }
+    } 
   }
 
   public boolean reOpenAllRegions(List<HRegionInfo> regions) throws IOException {
@@ -165,29 +161,6 @@ public abstract class TableEventHandler 
     return done;
   }
 
-  /**
-   * Table modifications are processed asynchronously, but provide an API for
-   * you to query their status.
-   *
-   * @throws IOException
-   */
-  public synchronized void waitForPersist() throws IOException {
-    if (!persistedToZk) {
-      try {
-        wait();
-      } catch (InterruptedException ie) {
-        throw (IOException) new InterruptedIOException().initCause(ie);
-      }
-      assert persistedToZk;
-    }
-  }
-
-  private synchronized void setPersist() {
-    if (!persistedToZk) {
-      persistedToZk = true;
-      notify();
-    }
-  }
 
   /**
    * Gets a TableDescriptor from the masterServices.  Can Throw exceptions.

Modified: hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/handler/TableModifyFamilyHandler.java
URL: http://svn.apache.org/viewvc/hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/handler/TableModifyFamilyHandler.java?rev=1446147&r1=1446146&r2=1446147&view=diff
==============================================================================
--- hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/handler/TableModifyFamilyHandler.java (original)
+++ hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/master/handler/TableModifyFamilyHandler.java Thu Feb 14 12:58:12 2013
@@ -55,11 +55,8 @@ public class TableModifyFamilyHandler ex
     if (cpHost != null) {
       cpHost.preModifyColumnHandler(this.tableName, this.familyDesc);
     }
-    // Update table descriptor in HDFS
-    HTableDescriptor htd =
-      this.masterServices.getMasterFileSystem().modifyColumn(tableName, familyDesc);
-    // Update in-memory descriptor cache
-    this.masterServices.getTableDescriptors().add(htd);
+    // Update table descriptor
+    this.masterServices.getMasterFileSystem().modifyColumn(tableName, familyDesc);
     if (cpHost != null) {
       cpHost.postModifyColumnHandler(this.tableName, this.familyDesc);
     }

Modified: hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/monitoring/LogMonitoring.java
URL: http://svn.apache.org/viewvc/hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/monitoring/LogMonitoring.java?rev=1446147&r1=1446146&r2=1446147&view=diff
==============================================================================
--- hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/monitoring/LogMonitoring.java (original)
+++ hbase/branches/hbase-7290v2/hbase-server/src/main/java/org/apache/hadoop/hbase/monitoring/LogMonitoring.java Thu Feb 14 12:58:12 2013
@@ -79,17 +79,18 @@ public abstract class LogMonitoring {
   private static void dumpTailOfLog(File f, PrintWriter out, long tailKb)
       throws IOException {
     FileInputStream fis = new FileInputStream(f);
+    BufferedReader r = null;
     try {
       FileChannel channel = fis.getChannel();
       channel.position(Math.max(0, channel.size() - tailKb*1024));
-      BufferedReader r = new BufferedReader(
-          new InputStreamReader(fis));
+      r = new BufferedReader(new InputStreamReader(fis));
       r.readLine(); // skip the first partial line
       String line;
       while ((line = r.readLine()) != null) {
         out.println(line);
       }
     } finally {
+      if (r != null) IOUtils.closeStream(r);
       IOUtils.closeStream(fis);
     }
   }



Mime
View raw message