hbase-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From st...@apache.org
Subject svn commit: r1208290 - in /hbase/branches/0.90: ./ src/main/java/org/apache/hadoop/hbase/io/ src/main/java/org/apache/hadoop/hbase/ipc/ src/main/java/org/apache/hadoop/hbase/master/ src/main/java/org/apache/hadoop/hbase/regionserver/ src/main/java/org/...
Date Wed, 30 Nov 2011 05:43:30 GMT
Author: stack
Date: Wed Nov 30 05:43:28 2011
New Revision: 1208290

URL: http://svn.apache.org/viewvc?rev=1208290&view=rev
Log:
HBASE-4352 Apply version of hbase-4015 to branch

Modified:
    hbase/branches/0.90/CHANGES.txt
    hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/io/HbaseObjectWritable.java
    hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/ipc/HRegionInterface.java
    hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/master/AssignmentManager.java
    hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/master/CatalogJanitor.java
    hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/master/HMaster.java
    hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/master/ServerManager.java
    hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/regionserver/HRegionServer.java
    hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/regionserver/handler/OpenMetaHandler.java
    hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/regionserver/handler/OpenRegionHandler.java
    hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/regionserver/handler/OpenRootHandler.java
    hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/zookeeper/ZKAssign.java
    hbase/branches/0.90/src/test/java/org/apache/hadoop/hbase/master/TestCatalogJanitor.java
    hbase/branches/0.90/src/test/java/org/apache/hadoop/hbase/master/TestZKBasedOpenCloseRegion.java

Modified: hbase/branches/0.90/CHANGES.txt
URL: http://svn.apache.org/viewvc/hbase/branches/0.90/CHANGES.txt?rev=1208290&r1=1208289&r2=1208290&view=diff
==============================================================================
--- hbase/branches/0.90/CHANGES.txt (original)
+++ hbase/branches/0.90/CHANGES.txt Wed Nov 30 05:43:28 2011
@@ -129,6 +129,7 @@ Release 0.90.5 - Unreleased
                (Jonathan Hsieh)
    HBASE-4806  Fix logging message in HbaseObjectWritable
                (Jonathan Hsieh via todd)
+   HBASE-4352  Apply version of hbase-4015 to branch
 
 
   NEW FEATURE

Modified: hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/io/HbaseObjectWritable.java
URL: http://svn.apache.org/viewvc/hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/io/HbaseObjectWritable.java?rev=1208290&r1=1208289&r2=1208290&view=diff
==============================================================================
--- hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/io/HbaseObjectWritable.java (original)
+++ hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/io/HbaseObjectWritable.java Wed Nov 30 05:43:28 2011
@@ -79,6 +79,7 @@ import org.apache.hadoop.hbase.filter.Va
 import org.apache.hadoop.hbase.filter.WhileMatchFilter;
 import org.apache.hadoop.hbase.filter.WritableByteArrayComparable;
 import org.apache.hadoop.hbase.regionserver.HRegion;
+import org.apache.hadoop.hbase.regionserver.RegionOpeningState;
 import org.apache.hadoop.hbase.regionserver.wal.HLog;
 import org.apache.hadoop.hbase.regionserver.wal.HLogKey;
 import org.apache.hadoop.hbase.util.Bytes;
@@ -213,6 +214,8 @@ public class HbaseObjectWritable impleme
     // serializable
     addToMap(Serializable.class, code++);
 
+    addToMap(RegionOpeningState.class, code++);
+
   }
 
   private Class<?> declaredClass;

Modified: hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/ipc/HRegionInterface.java
URL: http://svn.apache.org/viewvc/hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/ipc/HRegionInterface.java?rev=1208290&r1=1208289&r2=1208290&view=diff
==============================================================================
--- hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/ipc/HRegionInterface.java (original)
+++ hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/ipc/HRegionInterface.java Wed Nov 30 05:43:28 2011
@@ -38,6 +38,7 @@ import org.apache.hadoop.hbase.client.Mu
 import org.apache.hadoop.hbase.client.Put;
 import org.apache.hadoop.hbase.client.Result;
 import org.apache.hadoop.hbase.client.Scan;
+import org.apache.hadoop.hbase.regionserver.RegionOpeningState;
 import org.apache.hadoop.hbase.regionserver.wal.HLog;
 import org.apache.hadoop.hbase.util.Pair;
 import org.apache.hadoop.ipc.RemoteException;
@@ -312,10 +313,37 @@ public interface HRegionInterface extend
 
   /**
    * Opens the specified region.
-   * @param region region to open
+   * 
+   * @param region
+   *          region to open
+   * @param versionOfOfflineNode
+   *          the version of znode to compare when RS transitions the znode from
+   *          OFFLINE state.
+   * @return RegionOpeningState 
+   *         OPENED -         if region open request was successful.
+   *         ALREADY_OPENED - if the region was already opened. 
+   *         FAILED_OPENING - if region opening failed.
+   * @throws IOException
+   */
+  public RegionOpeningState openRegion(final HRegionInfo region) throws IOException;
+  
+  /**
+   * Opens the specified region.
+   * 
+   * @param region
+   *          region to open
+   * @param versionOfOfflineNode
+   *          the version of znode to compare when RS transitions the znode from
+   *          OFFLINE state.
+   * @return RegionOpeningState 
+   *         OPENED -         if region open request was successful.
+   *         ALREADY_OPENED - if the region was already opened. 
+   *         FAILED_OPENING - if region opening failed.
    * @throws IOException
    */
-  public void openRegion(final HRegionInfo region) throws IOException;
+  public RegionOpeningState openRegion(HRegionInfo region,
+      int versionOfOfflineNode) throws IOException;
+    
 
   /**
    * Opens the specified regions.

Modified: hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/master/AssignmentManager.java
URL: http://svn.apache.org/viewvc/hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/master/AssignmentManager.java?rev=1208290&r1=1208289&r2=1208290&view=diff
==============================================================================
--- hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/master/AssignmentManager.java (original)
+++ hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/master/AssignmentManager.java Wed Nov 30 05:43:28 2011
@@ -36,6 +36,7 @@ import java.util.TreeMap;
 import java.util.TreeSet;
 import java.util.concurrent.ConcurrentSkipListMap;
 import java.util.concurrent.ConcurrentSkipListSet;
+import java.util.concurrent.Executors;
 import java.util.concurrent.atomic.AtomicInteger;
 
 import org.apache.commons.logging.Log;
@@ -62,6 +63,7 @@ import org.apache.hadoop.hbase.master.ha
 import org.apache.hadoop.hbase.master.handler.DisableTableHandler;
 import org.apache.hadoop.hbase.master.handler.OpenedRegionHandler;
 import org.apache.hadoop.hbase.master.handler.ServerShutdownHandler;
+import org.apache.hadoop.hbase.regionserver.RegionOpeningState;
 import org.apache.hadoop.hbase.util.Bytes;
 import org.apache.hadoop.hbase.util.Pair;
 import org.apache.hadoop.hbase.util.Threads;
@@ -138,6 +140,9 @@ public class AssignmentManager extends Z
     new TreeMap<HRegionInfo,HServerInfo>();
 
   private final ExecutorService executorService;
+  
+  // Thread pool executor service for timeout monitor
+  private java.util.concurrent.ExecutorService threadPoolExecutorService;
 
   /**
    * Constructs a new assignment manager.
@@ -166,6 +171,7 @@ public class AssignmentManager extends Z
     this.zkTable = new ZKTable(this.master.getZooKeeper());
     this.maximumAssignmentAttempts =
       this.master.getConfiguration().getInt("hbase.assignment.maximum.attempts", 10);
+    this.threadPoolExecutorService = Executors.newCachedThreadPool();
   }
 
   /**
@@ -361,8 +367,19 @@ public class AssignmentManager extends Z
 
         // Just insert region into RIT
         // If this never updates the timeout will trigger new assignment
-        regionsInTransition.put(encodedRegionName, new RegionState(
-            regionInfo, RegionState.State.OPENING, data.getStamp()));
+        if (regionInfo.isMetaRegion() || regionInfo.isRootRegion()) {
+          regionsInTransition.put(encodedRegionName, new RegionState(
+              regionInfo, RegionState.State.OPENING, data.getStamp()));
+          // If ROOT or .META. table is waiting for timeout monitor to assign
+          // it may take lot of time when the assignment.timeout.period is
+          // the default value which may be very long.  We will not be able
+          // to serve any request during this time.
+          // So we will assign the ROOT and .META. region immediately.
+          processOpeningState(regionInfo);
+          break;
+        }
+        regionsInTransition.put(encodedRegionName, new RegionState(regionInfo,
+            RegionState.State.OPENING, data.getStamp()));
         break;
 
       case RS_ZK_REGION_OPENED:
@@ -854,26 +871,40 @@ public class AssignmentManager extends Z
   public void assign(HRegionInfo region, boolean setOfflineInZK) {
     assign(region, setOfflineInZK, false);
   }
-
   public void assign(HRegionInfo region, boolean setOfflineInZK,
       boolean forceNewPlan) {
+    assign(region, setOfflineInZK, forceNewPlan, false);
+  }
+
+  /**
+   * @param region
+   * @param setOfflineInZK
+   * @param forceNewPlan
+   * @param hijack - true new assignment is needed, false otherwise
+   */
+  public void assign(HRegionInfo region, boolean setOfflineInZK,
+      boolean forceNewPlan, boolean hijack) {
+    // If hijack is true do not call disableRegionIfInRIT as
+    // we have not yet moved the znode to OFFLINE state.
+    if (!hijack && isDisabledorDisablingRegionInRIT(region)) {
+      return;
+    }
+    RegionState state = addToRegionsInTransition(region, hijack);
+    synchronized (state) {
+      assign(region, state, setOfflineInZK, forceNewPlan, hijack);
+    }
+  }
+
+  private boolean isDisabledorDisablingRegionInRIT(HRegionInfo region) {
     String tableName = region.getTableDesc().getNameAsString();
     boolean disabled = this.zkTable.isDisabledTable(tableName);
     if (disabled || this.zkTable.isDisablingTable(tableName)) {
       LOG.info("Table " + tableName + (disabled? " disabled;": " disabling;") +
         " skipping assign of " + region.getRegionNameAsString());
       offlineDisabledRegion(region);
-      return;
-    }
-    if (this.serverManager.isClusterShutdown()) {
-      LOG.info("Cluster shutdown is set; skipping assign of " +
-        region.getRegionNameAsString());
-      return;
-    }
-    RegionState state = addToRegionsInTransition(region);
-    synchronized (state) {
-      assign(state, setOfflineInZK, forceNewPlan);
+      return true;
     }
+    return false;
   }
 
   /**
@@ -1008,11 +1039,13 @@ public class AssignmentManager extends Z
 
   /**
    * @param region
-   * @return
+   * @param hijack
+   * @return The current RegionState
    */
-  private RegionState addToRegionsInTransition(final HRegionInfo region) {
+  private RegionState addToRegionsInTransition(final HRegionInfo region,
+      boolean hijack) {
     synchronized (regionsInTransition) {
-      return forceRegionStateToOffline(region);
+      return forceRegionStateToOffline(region, hijack);
     }
   }
 
@@ -1023,14 +1056,31 @@ public class AssignmentManager extends Z
    * @return Amended RegionState.
    */
   private RegionState forceRegionStateToOffline(final HRegionInfo region) {
+    return forceRegionStateToOffline(region, false);
+  }
+  /**
+   * Sets regions {@link RegionState} to {@link RegionState.State#OFFLINE}.
+   * Caller must hold lock on this.regionsInTransition.
+   * @param region
+   * @param hijack
+   * @return Amended RegionState.
+   */
+  private RegionState forceRegionStateToOffline(final HRegionInfo region,
+      boolean hijack) {
     String encodedName = region.getEncodedName();
     RegionState state = this.regionsInTransition.get(encodedName);
     if (state == null) {
       state = new RegionState(region, RegionState.State.OFFLINE);
       this.regionsInTransition.put(encodedName, state);
     } else {
-      LOG.debug("Forcing OFFLINE; was=" + state);
-      state.update(RegionState.State.OFFLINE);
+      // If we are reassigning the node do not force in-memory state to OFFLINE.
+      // Based on the znode state we will decide if to change
+      // in-memory state to OFFLINE or not. It will
+      // be done before setting the znode to OFFLINE state.
+      if (!hijack) {
+        LOG.debug("Forcing OFFLINE; was=" + state);
+        state.update(RegionState.State.OFFLINE);
+      }
     }
     return state;
   }
@@ -1040,11 +1090,27 @@ public class AssignmentManager extends Z
    * @param state
    * @param setOfflineInZK
    * @param forceNewPlan
+   * @param hijack
    */
-  private void assign(final RegionState state, final boolean setOfflineInZK,
-      final boolean forceNewPlan) {
+  private void assign(final HRegionInfo region, final RegionState state,
+      final boolean setOfflineInZK, final boolean forceNewPlan, boolean hijack) {
     for (int i = 0; i < this.maximumAssignmentAttempts; i++) {
-      if (setOfflineInZK && !setOfflineInZooKeeper(state)) return;
+      int versionOfOfflineNode = -1;
+      if (setOfflineInZK) {
+        // get the version of the znode after setting it to OFFLINE.
+        // versionOfOfflineNode will be -1 if the znode was not set to OFFLINE
+        versionOfOfflineNode = setOfflineInZooKeeper(state, hijack);
+        if (versionOfOfflineNode != -1) {
+          if (isDisabledorDisablingRegionInRIT(region)) {
+            return;
+          }
+        }
+      }
+
+      if (setOfflineInZK && versionOfOfflineNode == -1) {
+        return;
+      }
+            
       if (this.master.isStopped()) {
         LOG.debug("Server stopped; skipping assign of " + state);
         return;
@@ -1062,7 +1128,34 @@ public class AssignmentManager extends Z
         // Transition RegionState to PENDING_OPEN
         state.update(RegionState.State.PENDING_OPEN);
         // Send OPEN RPC. This can fail if the server on other end is is not up.
-        serverManager.sendRegionOpen(plan.getDestination(), state.getRegion());
+        // Pass the version that was obtained while setting the node to OFFLINE.
+        RegionOpeningState regionOpenState = serverManager.sendRegionOpen(plan
+            .getDestination(), state.getRegion(), versionOfOfflineNode);
+        if (regionOpenState == RegionOpeningState.ALREADY_OPENED) {
+          // Remove region from in-memory transition and unassigned node from ZK
+          // While trying to enable the table the regions of the table were
+          // already enabled.
+          String encodedRegionName = state.getRegion()
+              .getEncodedName();
+          try {
+            ZKAssign.deleteOfflineNode(master.getZooKeeper(), encodedRegionName);
+          } catch (KeeperException.NoNodeException e) {
+            if(LOG.isDebugEnabled()){
+              LOG.debug("The unassigned node "+encodedRegionName+" doesnot exist.");
+            }
+          } catch (KeeperException e) {
+            master.abort(
+                "Error deleting OFFLINED node in ZK for transition ZK node ("
+                    + encodedRegionName + ")", e);
+          }
+          synchronized (this.regionsInTransition) {
+            this.regionsInTransition.remove(plan.getRegionInfo()
+                .getEncodedName());
+          }
+          synchronized (this.regions) {
+            this.regions.put(plan.getRegionInfo(), plan.getDestination());
+          }
+        }
         break;
       } catch (Throwable t) {
         LOG.warn("Failed assignment of " +
@@ -1095,29 +1188,53 @@ public class AssignmentManager extends Z
   /**
    * Set region as OFFLINED up in zookeeper
    * @param state
-   * @return True if we succeeded, false otherwise (State was incorrect or failed
-   * updating zk).
-   */
-  boolean setOfflineInZooKeeper(final RegionState state) {
-    if (!state.isClosed() && !state.isOffline()) {
+   * @param hijack
+   *          - true if needs to be hijacked and reassigned, false otherwise.
+   * @return the version of the offline node if setting of the OFFLINE node was
+   *         successful, -1 otherwise.
+   */
+  int setOfflineInZooKeeper(final RegionState state, boolean hijack) {
+    // In case of reassignment the current state in memory need not be
+    // OFFLINE. 
+    if (!hijack && !state.isClosed() && !state.isOffline()) {
         new RuntimeException("Unexpected state trying to OFFLINE; " + state);
       this.master.abort("Unexpected state trying to OFFLINE; " + state,
         new IllegalStateException());
-      return false;
+      return -1;
     }
-    state.update(RegionState.State.OFFLINE);
+    boolean allowZNodeCreation = false;
+    // Under reassignment if the current state is PENDING_OPEN
+    // or OPENING then refresh the in-memory state to PENDING_OPEN. This is
+    // important because if the region was in
+    // RS_OPENING state for a long time the master will try to force the znode
+    // to OFFLINE state meanwhile the RS could have opened the corresponding
+    // region and the state in znode will be RS_ZK_REGION_OPENED.
+    // For all other cases we can change the in-memory state to OFFLINE.
+    if (hijack
+        && (state.getState().equals(RegionState.State.PENDING_OPEN) || state
+            .getState().equals(RegionState.State.OPENING))) {
+      state.update(RegionState.State.PENDING_OPEN);
+      allowZNodeCreation = false;
+    } else {
+      state.update(RegionState.State.OFFLINE);
+      allowZNodeCreation = true;
+    }
+    int versionOfOfflineNode = -1;
     try {
-      if(!ZKAssign.createOrForceNodeOffline(master.getZooKeeper(),
-          state.getRegion(), master.getServerName())) {
+      // get the version after setting the znode to OFFLINE
+      versionOfOfflineNode = ZKAssign.createOrForceNodeOffline(master
+          .getZooKeeper(), state.getRegion(), this.master.getServerName(),
+          hijack, allowZNodeCreation);
+      if (versionOfOfflineNode == -1) {
         LOG.warn("Attempted to create/force node into OFFLINE state before " +
           "completing assignment but failed to do so for " + state);
-        return false;
+        return -1;
       }
     } catch (KeeperException e) {
       master.abort("Unexpected ZK exception creating/setting node OFFLINE", e);
-      return false;
+      return -1;
     }
-    return true;
+    return versionOfOfflineNode;
   }
 
   /**
@@ -1848,135 +1965,134 @@ public class AssignmentManager extends Z
     protected void chore() {
       // If bulkAssign in progress, suspend checks
       if (this.bulkAssign) return;
-      List<HRegionInfo> unassigns = new ArrayList<HRegionInfo>();
-      Map<HRegionInfo, Boolean> assigns =
-        new HashMap<HRegionInfo, Boolean>();
       synchronized (regionsInTransition) {
         // Iterate all regions in transition checking for time outs
         long now = System.currentTimeMillis();
         for (RegionState regionState : regionsInTransition.values()) {
           if (regionState.getStamp() + timeout <= now) {
-            HRegionInfo regionInfo = regionState.getRegion();
-            LOG.info("Regions in transition timed out:  " + regionState);
-            // Expired!  Do a retry.
-            switch (regionState.getState()) {
-              case CLOSED:
-                LOG.info("Region " + regionInfo.getEncodedName() +
-                  " has been CLOSED for too long, waiting on queued " +
-                  "ClosedRegionHandler to run or server shutdown");
-                // Update our timestamp.
-                synchronized(regionState) {
-                  regionState.update(regionState.getState());
-                }
-                break;
-              case OFFLINE:
-                LOG.info("Region has been OFFLINE for too long, " +
-                  "reassigning " + regionInfo.getRegionNameAsString() +
-                  " to a random server");
-                assigns.put(regionState.getRegion(), Boolean.FALSE);
-                break;
-              case PENDING_OPEN:
-                LOG.info("Region has been PENDING_OPEN for too " +
-                    "long, reassigning region=" +
-                    regionInfo.getRegionNameAsString());
-                assigns.put(regionState.getRegion(), Boolean.TRUE);
-                break;
-              case OPENING:
-                LOG.info("Region has been OPENING for too " +
-                  "long, reassigning region=" +
-                  regionInfo.getRegionNameAsString());
-                // Should have a ZK node in OPENING state
-                try {
-                  String node = ZKAssign.getNodeName(watcher,
-                      regionInfo.getEncodedName());
-                  Stat stat = new Stat();
-                  RegionTransitionData data = ZKAssign.getDataNoWatch(watcher,
-                      node, stat);
-                  if (data == null) {
-                    LOG.warn("Data is null, node " + node + " no longer exists");
-                    break;
-                  }
-                  if (data.getEventType() == EventType.RS_ZK_REGION_OPENED) {
-                    LOG.debug("Region has transitioned to OPENED, allowing " +
-                        "watched event handlers to process");
-                    break;
-                  } else if (data.getEventType() !=
-                      EventType.RS_ZK_REGION_OPENING) {
-                    LOG.warn("While timing out a region in state OPENING, " +
-                        "found ZK node in unexpected state: " +
-                        data.getEventType());
-                    break;
-                  }
-                  // Attempt to transition node into OFFLINE
-                  try {
-                    data = new RegionTransitionData(
-                      EventType.M_ZK_REGION_OFFLINE, regionInfo.getRegionName(),
-                      master.getServerName());
-                    if (ZKUtil.setData(watcher, node, data.getBytes(),
-                        stat.getVersion())) {
-                      // Node is now OFFLINE, let's trigger another assignment
-                      ZKUtil.getDataAndWatch(watcher, node); // re-set the watch
-                      LOG.info("Successfully transitioned region=" +
-                          regionInfo.getRegionNameAsString() + " into OFFLINE" +
-                          " and forcing a new assignment");
-                      assigns.put(regionState.getRegion(), Boolean.TRUE);
-                    }
-                  } catch (KeeperException.NoNodeException nne) {
-                    // Node did not exist, can't time this out
-                  }
-                } catch (KeeperException ke) {
-                  LOG.error("Unexpected ZK exception timing out CLOSING region",
-                      ke);
-                  break;
-                }
-                break;
-              case OPEN:
-                LOG.error("Region has been OPEN for too long, " +
-                "we don't know where region was opened so can't do anything");
-                synchronized(regionState) {
-                  regionState.update(regionState.getState());
-                }
-                break;
-
-              case PENDING_CLOSE:
-                LOG.info("Region has been PENDING_CLOSE for too " +
-                    "long, running forced unassign again on region=" +
-                    regionInfo.getRegionNameAsString());
-                  try {
-                    // If the server got the RPC, it will transition the node
-                    // to CLOSING, so only do something here if no node exists
-                    if (!ZKUtil.watchAndCheckExists(watcher,
-                      ZKAssign.getNodeName(watcher, regionInfo.getEncodedName()))) {
-                      // Queue running of an unassign -- do actual unassign
-                      // outside of the regionsInTransition lock.
-                      unassigns.add(regionInfo);
-                    }
-                  } catch (NoNodeException e) {
-                    LOG.debug("Node no longer existed so not forcing another " +
-                      "unassignment");
-                  } catch (KeeperException e) {
-                    LOG.warn("Unexpected ZK exception timing out a region " +
-                      "close", e);
-                  }
-                  break;
-              case CLOSING:
-                LOG.info("Region has been CLOSING for too " +
-                  "long, this should eventually complete or the server will " +
-                  "expire, doing nothing");
-                break;
-            }
+            actOnTimeOut(regionState);
           }
         }
       }
-      // Finish the work for regions in PENDING_CLOSE state
-      for (HRegionInfo hri: unassigns) {
-        unassign(hri, true);
+    }
+
+    private void actOnTimeOut(RegionState regionState) {
+      HRegionInfo regionInfo = regionState.getRegion();
+      LOG.info("Regions in transition timed out:  " + regionState);
+      // Expired!  Do a retry.
+      switch (regionState.getState()) {
+        case CLOSED:
+          LOG.info("Region " + regionInfo.getEncodedName() +
+            " has been CLOSED for too long, waiting on queued " +
+            "ClosedRegionHandler to run or server shutdown");
+          // Update our timestamp.
+          synchronized(regionState) {
+            regionState.update(regionState.getState());
+          }
+          break;
+        case OFFLINE:
+          LOG.info("Region has been OFFLINE for too long, " +
+            "reassigning " + regionInfo.getRegionNameAsString() +
+            " to a random server");
+          invokeAssign(regionInfo);
+          break;
+        case PENDING_OPEN:
+          LOG.info("Region has been PENDING_OPEN for too " +
+              "long, reassigning region=" +
+              regionInfo.getRegionNameAsString());
+          invokeAssign(regionInfo);
+          break;
+        case OPENING:
+          processOpeningState(regionInfo);
+          break;
+        case OPEN:
+          LOG.error("Region has been OPEN for too long, " +
+          "we don't know where region was opened so can't do anything");
+          synchronized(regionState) {
+            regionState.update(regionState.getState());
+          }
+          break;
+
+        case PENDING_CLOSE:
+          LOG.info("Region has been PENDING_CLOSE for too " +
+              "long, running forced unassign again on region=" +
+              regionInfo.getRegionNameAsString());
+            try {
+              // If the server got the RPC, it will transition the node
+              // to CLOSING, so only do something here if no node exists
+              if (!ZKUtil.watchAndCheckExists(watcher,
+                ZKAssign.getNodeName(watcher, regionInfo.getEncodedName()))) {
+                // Queue running of an unassign -- do actual unassign
+                // outside of the regionsInTransition lock.
+                invokeUnassign(regionInfo);
+              }
+            } catch (NoNodeException e) {
+              LOG.debug("Node no longer existed so not forcing another " +
+                "unassignment");
+            } catch (KeeperException e) {
+              LOG.warn("Unexpected ZK exception timing out a region " +
+                "close", e);
+            }
+            break;
+        case CLOSING:
+          LOG.info("Region has been CLOSING for too " +
+            "long, this should eventually complete or the server will " +
+            "expire, doing nothing");
+          break;
+      }
+    }
+  }
+  
+  private void processOpeningState(HRegionInfo regionInfo) {
+    LOG.info("Region has been OPENING for too " + "long, reassigning region=" + 
+      regionInfo.getRegionNameAsString());
+    // Should have a ZK node in OPENING state
+    try {
+      String node = ZKAssign.getNodeName(watcher, regionInfo.getEncodedName());
+      Stat stat = new Stat();
+      RegionTransitionData data = ZKAssign.getDataNoWatch(watcher, node, stat);
+      if (data == null) {
+        LOG.warn("Data is null, node " + node + " no longer exists");
+        return;
+      }
+      if (data.getEventType() == EventType.RS_ZK_REGION_OPENED) {
+        LOG.debug("Region has transitioned to OPENED, allowing " + 
+          "watched event handlers to process");
+        return;
+      } else if (data.getEventType() != EventType.RS_ZK_REGION_OPENING) {
+        LOG.warn("While timing out a region in state OPENING, " +
+          "found ZK node in unexpected state: " + data.getEventType());
+        return;
       }
-      for (Map.Entry<HRegionInfo, Boolean> e: assigns.entrySet()){
-        assign(e.getKey(), false, e.getValue());
+      // Attempt to transition node into OFFLINE
+      try {
+        data = new RegionTransitionData(EventType.M_ZK_REGION_OFFLINE,
+            regionInfo.getRegionName(), master.getServerName());
+        if (ZKUtil.setData(watcher, node, data.getBytes(), stat.getVersion())) {
+          // Node is now OFFLINE, let's trigger another assignment
+          ZKUtil.getDataAndWatch(watcher, node); // re-set the watch
+          LOG.info("Successfully transitioned region=" +
+            regionInfo.getRegionNameAsString() + " into OFFLINE" +
+            " and forcing a new assignment");
+          invokeAssign(regionInfo);
+        }
+      } catch (KeeperException.NoNodeException nne) {
+        // Node did not exist, can't time this out
       }
+    } catch (KeeperException ke) {
+      LOG.error("Unexpected ZK exception timing out CLOSING region", ke);
+      return;
     }
   }
+  
+    
+  private void invokeAssign(HRegionInfo regionInfo) {
+    threadPoolExecutorService.submit(new AssignCallable(this, regionInfo));
+  }
+
+  private void invokeUnassign(HRegionInfo regionInfo) {
+    threadPoolExecutorService.submit(new UnAssignCallable(this, regionInfo));
+  }
 
   /**
    * Process shutdown server removing any assignments.
@@ -2266,4 +2382,12 @@ public class AssignmentManager extends Z
   public boolean isServerOnline(String serverName) {
     return this.serverManager.isServerOnline(serverName);
   }
+  /**
+   * Shutdown the threadpool executor service
+   */
+  public void shutdown() {
+    if (null != threadPoolExecutorService) {
+      this.threadPoolExecutorService.shutdown();
+    }
+  }
 }

Modified: hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/master/CatalogJanitor.java
URL: http://svn.apache.org/viewvc/hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/master/CatalogJanitor.java?rev=1208290&r1=1208289&r2=1208290&view=diff
==============================================================================
--- hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/master/CatalogJanitor.java (original)
+++ hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/master/CatalogJanitor.java Wed Nov 30 05:43:28 2011
@@ -157,6 +157,7 @@ class CatalogJanitor extends Chore {
     }
   }
 
+
   /**
    * Get HRegionInfo from passed Map of row values.
    * @param result Map to do lookup in.

Modified: hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/master/HMaster.java
URL: http://svn.apache.org/viewvc/hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/master/HMaster.java?rev=1208290&r1=1208289&r2=1208290&view=diff
==============================================================================
--- hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/master/HMaster.java (original)
+++ hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/master/HMaster.java Wed Nov 30 05:43:28 2011
@@ -965,6 +965,7 @@ implements HMasterInterface, HMasterRegi
 
   @Override
   public void shutdown() {
+    this.assignmentManager.shutdown();
     this.serverManager.shutdownCluster();
     try {
       this.clusterStatusTracker.setClusterDown();

Modified: hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/master/ServerManager.java
URL: http://svn.apache.org/viewvc/hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/master/ServerManager.java?rev=1208290&r1=1208289&r2=1208290&view=diff
==============================================================================
--- hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/master/ServerManager.java (original)
+++ hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/master/ServerManager.java Wed Nov 30 05:43:28 2011
@@ -49,6 +49,7 @@ import org.apache.hadoop.hbase.ipc.HRegi
 import org.apache.hadoop.hbase.master.handler.MetaServerShutdownHandler;
 import org.apache.hadoop.hbase.master.handler.ServerShutdownHandler;
 import org.apache.hadoop.hbase.master.metrics.MasterMetrics;
+import org.apache.hadoop.hbase.regionserver.RegionOpeningState;
 import org.apache.hadoop.hbase.regionserver.Leases.LeaseStillHeldException;
 
 /**
@@ -558,16 +559,20 @@ public class ServerManager {
    * <p>
    * @param server server to open a region
    * @param region region to open
+   * @param versionOfOfflineNode that needs to be present in the offline node
+   * when RS tries to change the state from OFFLINE to other states.
    */
-  public void sendRegionOpen(HServerInfo server, HRegionInfo region)
+  public RegionOpeningState sendRegionOpen(final HServerInfo server,
+      HRegionInfo region, int versionOfOfflineNode)
   throws IOException {
     HRegionInterface hri = getServerConnection(server);
     if (hri == null) {
-      LOG.warn("Attempting to send OPEN RPC to server " + server.getServerName()
-          + " failed because no RPC connection found to this server");
-      return;
+      LOG.warn("Attempting to send OPEN RPC to server " + server.toString() +
+        " failed because no RPC connection found to this server");
+      return RegionOpeningState.FAILED_OPENING;
     }
-    hri.openRegion(region);
+    return (versionOfOfflineNode == -1) ? hri.openRegion(region) : hri
+        .openRegion(region, versionOfOfflineNode);
   }
 
   /**

Modified: hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/regionserver/HRegionServer.java
URL: http://svn.apache.org/viewvc/hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/regionserver/HRegionServer.java?rev=1208290&r1=1208289&r2=1208290&view=diff
==============================================================================
--- hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/regionserver/HRegionServer.java (original)
+++ hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/regionserver/HRegionServer.java Wed Nov 30 05:43:28 2011
@@ -67,6 +67,7 @@ import org.apache.hadoop.hbase.HRegionIn
 import org.apache.hadoop.hbase.HServerAddress;
 import org.apache.hadoop.hbase.HServerInfo;
 import org.apache.hadoop.hbase.HServerLoad;
+import org.apache.hadoop.hbase.HTableDescriptor;
 import org.apache.hadoop.hbase.KeyValue;
 import org.apache.hadoop.hbase.MasterAddressTracker;
 import org.apache.hadoop.hbase.NotServingRegionException;
@@ -2124,24 +2125,40 @@ public class HRegionServer implements HR
 
   // Region open/close direct RPCs
 
+  @QosPriority(priority=HIGH_QOS)
+  public RegionOpeningState openRegion(HRegionInfo region)
+  throws IOException {
+    return openRegion(region, -1);
+  }
   @Override
   @QosPriority(priority=HIGH_QOS)
-  public void openRegion(HRegionInfo region)
+  public RegionOpeningState openRegion(HRegionInfo region, int versionOfOfflineNode)
   throws IOException {
     if (this.regionsInTransitionInRS.contains(region.getEncodedNameAsBytes())) {
       throw new RegionAlreadyInTransitionException("open", region.getEncodedName());
     }
+    HRegion onlineRegion = this.getFromOnlineRegions(region.getEncodedName());
+    if (null != onlineRegion) {
+      LOG.warn("Attempted open of " + region.getEncodedName()
+          + " but already online on this server");
+      return RegionOpeningState.ALREADY_OPENED;
+    }
     LOG.info("Received request to open region: " +
       region.getRegionNameAsString());
     if (this.stopped) throw new RegionServerStoppedException();
     this.regionsInTransitionInRS.add(region.getEncodedNameAsBytes());
+    // Need to pass the expected version in the constructor.
     if (region.isRootRegion()) {
-      this.service.submit(new OpenRootHandler(this, this, region));
-    } else if(region.isMetaRegion()) {
-      this.service.submit(new OpenMetaHandler(this, this, region));
+      this.service.submit(new OpenRootHandler(this, this, region,
+          versionOfOfflineNode));
+    } else if (region.isMetaRegion()) {
+      this.service.submit(new OpenMetaHandler(this, this, region,
+          versionOfOfflineNode));
     } else {
-      this.service.submit(new OpenRegionHandler(this, this, region));
+      this.service.submit(new OpenRegionHandler(this, this, region,
+          versionOfOfflineNode));
     }
+    return RegionOpeningState.OPENED;
   }
 
   @Override

Modified: hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/regionserver/handler/OpenMetaHandler.java
URL: http://svn.apache.org/viewvc/hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/regionserver/handler/OpenMetaHandler.java?rev=1208290&r1=1208289&r2=1208290&view=diff
==============================================================================
--- hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/regionserver/handler/OpenMetaHandler.java (original)
+++ hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/regionserver/handler/OpenMetaHandler.java Wed Nov 30 05:43:28 2011
@@ -31,6 +31,12 @@ import org.apache.hadoop.hbase.regionser
 public class OpenMetaHandler extends OpenRegionHandler {
   public OpenMetaHandler(final Server server,
       final RegionServerServices rsServices, HRegionInfo regionInfo) {
-    super(server,rsServices,  regionInfo, EventType.M_RS_OPEN_META);
+    super(server,rsServices,  regionInfo, EventType.M_RS_OPEN_META, -1);
+  }
+  public OpenMetaHandler(final Server server,
+      final RegionServerServices rsServices, HRegionInfo regionInfo,
+      int versionOfOfflineNode) {
+    super(server, rsServices, regionInfo, EventType.M_RS_OPEN_META,
+        versionOfOfflineNode);
   }
 }

Modified: hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/regionserver/handler/OpenRegionHandler.java
URL: http://svn.apache.org/viewvc/hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/regionserver/handler/OpenRegionHandler.java?rev=1208290&r1=1208289&r2=1208290&view=diff
==============================================================================
--- hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/regionserver/handler/OpenRegionHandler.java (original)
+++ hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/regionserver/handler/OpenRegionHandler.java Wed Nov 30 05:43:28 2011
@@ -49,18 +49,27 @@ public class OpenRegionHandler extends E
   // the total open. We'll fail the open if someone hijacks our znode; we can
   // tell this has happened if version is not as expected.
   private volatile int version = -1;
+  //version of the offline node that was set by the master
+  private volatile int versionOfOfflineNode = -1;
 
   public OpenRegionHandler(final Server server,
       final RegionServerServices rsServices, HRegionInfo regionInfo) {
-    this(server, rsServices, regionInfo, EventType.M_RS_OPEN_REGION);
+    this(server, rsServices, regionInfo, EventType.M_RS_OPEN_REGION, -1);
+  }
+  public OpenRegionHandler(final Server server,
+      final RegionServerServices rsServices, HRegionInfo regionInfo,
+      int versionOfOfflineNode) {
+    this(server, rsServices, regionInfo, EventType.M_RS_OPEN_REGION,
+        versionOfOfflineNode);
   }
 
   protected OpenRegionHandler(final Server server,
       final RegionServerServices rsServices, final HRegionInfo regionInfo,
-      EventType eventType) {
+      EventType eventType, int versionOfOfflineNode) {
     super(server, eventType);
     this.rsServices = rsServices;
     this.regionInfo = regionInfo;
+    this.versionOfOfflineNode = versionOfOfflineNode;
   }
 
   public HRegionInfo getRegionInfo() {
@@ -92,7 +101,8 @@ public class OpenRegionHandler extends E
 
       // If fails, just return.  Someone stole the region from under us.
       // Calling transitionZookeeperOfflineToOpening initalizes this.version.
-      if (!transitionZookeeperOfflineToOpening(encodedName)) {
+      if (!transitionZookeeperOfflineToOpening(encodedName,
+          versionOfOfflineNode)) {
         LOG.warn("Region was hijacked? It no longer exists, encodedName=" +
           encodedName);
         return;
@@ -303,15 +313,18 @@ public class OpenRegionHandler extends E
    * Transition ZK node from OFFLINE to OPENING.
    * @param encodedName Name of the znode file (Region encodedName is the znode
    * name).
+   * @param versionOfOfflineNode - version Of OfflineNode that needs to be compared
+   * before changing the node's state from OFFLINE
    * @return True if successful transition.
    */
-  boolean transitionZookeeperOfflineToOpening(final String encodedName) {
+  boolean transitionZookeeperOfflineToOpening(final String encodedName,
+      int versionOfOfflineNode) {
     // TODO: should also handle transition from CLOSED?
     try {
       // Initialize the znode version.
-      this.version =
-        ZKAssign.transitionNodeOpening(server.getZooKeeper(),
-          regionInfo, server.getServerName());
+      this.version = ZKAssign.transitionNode(server.getZooKeeper(), regionInfo,
+          server.getServerName(), EventType.M_ZK_REGION_OFFLINE,
+          EventType.RS_ZK_REGION_OPENING, versionOfOfflineNode);
     } catch (KeeperException e) {
       LOG.error("Error transition from OFFLINE to OPENING for region=" +
         encodedName, e);

Modified: hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/regionserver/handler/OpenRootHandler.java
URL: http://svn.apache.org/viewvc/hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/regionserver/handler/OpenRootHandler.java?rev=1208290&r1=1208289&r2=1208290&view=diff
==============================================================================
--- hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/regionserver/handler/OpenRootHandler.java (original)
+++ hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/regionserver/handler/OpenRootHandler.java Wed Nov 30 05:43:28 2011
@@ -29,8 +29,15 @@ import org.apache.hadoop.hbase.regionser
  * This is executed after receiving an OPEN RPC from the master for root.
  */
 public class OpenRootHandler extends OpenRegionHandler {
+  
   public OpenRootHandler(final Server server,
       final RegionServerServices rsServices, HRegionInfo regionInfo) {
-    super(server, rsServices, regionInfo, EventType.M_RS_OPEN_ROOT);
+    super(server, rsServices, regionInfo, EventType.M_RS_OPEN_ROOT, -1);
+  }
+  public OpenRootHandler(final Server server,
+      final RegionServerServices rsServices, HRegionInfo regionInfo,
+      int versionOfOfflineNode) {
+    super(server, rsServices, regionInfo, EventType.M_RS_OPEN_ROOT,
+        versionOfOfflineNode);
   }
 }

Modified: hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/zookeeper/ZKAssign.java
URL: http://svn.apache.org/viewvc/hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/zookeeper/ZKAssign.java?rev=1208290&r1=1208289&r2=1208290&view=diff
==============================================================================
--- hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/zookeeper/ZKAssign.java (original)
+++ hbase/branches/0.90/src/main/java/org/apache/hadoop/hbase/zookeeper/ZKAssign.java Wed Nov 30 05:43:28 2011
@@ -211,7 +211,6 @@ public class ZKAssign {
     }
   }
 
-
   /**
    * Creates or force updates an unassigned node to the OFFLINE state for the
    * specified region.
@@ -227,11 +226,48 @@ public class ZKAssign {
    * @param zkw zk reference
    * @param region region to be created as offline
    * @param serverName server event originates from
+   * @return the version of the znode created in OFFLINE state, -1 if
+   *         unsuccessful.
    * @throws KeeperException if unexpected zookeeper exception
    * @throws KeeperException.NodeExistsException if node already exists
    */
-  public static boolean createOrForceNodeOffline(ZooKeeperWatcher zkw,
-      HRegionInfo region, String serverName)
+  public static int createOrForceNodeOffline(ZooKeeperWatcher zkw,
+      HRegionInfo region, String serverName) throws KeeperException {
+    return createOrForceNodeOffline(zkw, region, serverName, false, true);
+  }
+
+  /**
+   * Creates or force updates an unassigned node to the OFFLINE state for the
+   * specified region.
+   * <p>
+   * Attempts to create the node but if it exists will force it to transition to
+   * and OFFLINE state.
+   * <p>
+   * Sets a watcher on the unassigned region node if the method is successful.
+   * 
+   * <p>
+   * This method should be used when assigning a region.
+   * 
+   * @param zkw
+   *          zk reference
+   * @param region
+   *          region to be created as offline
+   * @param serverName
+   *          server event originates from
+   * @param hijack
+   *          - true if to be hijacked and reassigned, false otherwise
+   * @param allowCreation
+   *          - true if the node has to be created newly, false otherwise
+   * @throws KeeperException
+   *           if unexpected zookeeper exception
+   * @return the version of the znode created in OFFLINE state, -1 if
+   *         unsuccessful.
+   * @throws KeeperException.NodeExistsException
+   *           if node already exists
+   */
+  public static int createOrForceNodeOffline(ZooKeeperWatcher zkw,
+      HRegionInfo region, String serverName,
+      boolean hijack, boolean allowCreation)
   throws KeeperException {
     LOG.debug(zkw.prefix("Creating (or updating) unassigned node for " +
       region.getEncodedName() + " with OFFLINE state"));
@@ -239,14 +275,49 @@ public class ZKAssign {
         EventType.M_ZK_REGION_OFFLINE, region.getRegionName(), serverName);
     synchronized(zkw.getNodes()) {
       String node = getNodeName(zkw, region.getEncodedName());
+      Stat stat = new Stat();
       zkw.sync(node);
       zkw.getNodes().add(node);
       int version = ZKUtil.checkExists(zkw, node);
-      if(version == -1) {
-        ZKUtil.createAndWatch(zkw, node, data.getBytes());
+      if (version == -1) {
+        // While trying to transit a node to OFFLINE that was in previously in 
+        // OPENING state but before it could transit to OFFLINE state if RS had 
+        // opened the region then the Master deletes the assigned region znode. 
+        // In that case the znode will not exist. So we should not
+        // create the znode again which will lead to double assignment.
+        if (hijack && !allowCreation) {
+          return -1;
+        }
+        return ZKUtil.createAndWatch(zkw, node, data.getBytes());
       } else {
-        if (!ZKUtil.setData(zkw, node, data.getBytes(), version)) {
-          return false;
+        RegionTransitionData curDataInZNode = ZKAssign.getDataNoWatch(zkw, region
+            .getEncodedName(), stat);
+        // Do not move the node to OFFLINE if znode is in any of the following
+        // state.
+        // Because these are already executed states.
+        if (hijack && null != curDataInZNode) {
+          EventType eventType = curDataInZNode.getEventType();
+          if (eventType.equals(EventType.RS_ZK_REGION_CLOSING)
+              || eventType.equals(EventType.RS_ZK_REGION_CLOSED)
+              || eventType.equals(EventType.RS_ZK_REGION_OPENED)) {
+            return -1;
+          }
+        }
+
+        boolean setData = false;
+        try {
+          setData = ZKUtil.setData(zkw, node, data.getBytes(), version);
+          // Setdata throws KeeperException which aborts the Master. So we are
+          // catching it here.
+          // If just before setting the znode to OFFLINE if the RS has made any
+          // change to the
+          // znode state then we need to return -1.
+        } catch (KeeperException kpe) {
+          LOG.info("Version mismatch while setting the node to OFFLINE state.");
+          return -1;
+        }
+        if (!setData) {
+          return -1;
         } else {
           // We successfully forced to OFFLINE, reset watch and handle if
           // the state changed in between our set and the watch
@@ -254,12 +325,12 @@ public class ZKAssign {
             ZKAssign.getData(zkw, region.getEncodedName());
           if (curData.getEventType() != data.getEventType()) {
             // state changed, need to process
-            return false;
+            return -1;
           }
         }
       }
+      return stat.getVersion() + 1;
     }
-    return true;
   }
 
   /**
@@ -720,6 +791,18 @@ public class ZKAssign {
         "the node existed but was version " + stat.getVersion() +
         " not the expected version " + expectedVersion));
         return -1;
+    } else if (beginState.equals(EventType.M_ZK_REGION_OFFLINE)
+        && endState.equals(EventType.RS_ZK_REGION_OPENING)
+        && expectedVersion == -1 && stat.getVersion() != 0) {
+      // the below check ensures that double assignment doesnot happen.
+      // When the node is created for the first time then the expected version
+      // that is passed will be -1 and the version in znode will be 0.
+      // In all other cases the version in znode will be > 0.
+      LOG.warn(zkw.prefix("Attempt to transition the " + "unassigned node for "
+          + encoded + " from " + beginState + " to " + endState + " failed, "
+          + "the node existed but was version " + stat.getVersion()
+          + " not the expected version " + expectedVersion));
+      return -1;
     }
 
     // Verify it is in expected state

Modified: hbase/branches/0.90/src/test/java/org/apache/hadoop/hbase/master/TestCatalogJanitor.java
URL: http://svn.apache.org/viewvc/hbase/branches/0.90/src/test/java/org/apache/hadoop/hbase/master/TestCatalogJanitor.java?rev=1208290&r1=1208289&r2=1208290&view=diff
==============================================================================
--- hbase/branches/0.90/src/test/java/org/apache/hadoop/hbase/master/TestCatalogJanitor.java (original)
+++ hbase/branches/0.90/src/test/java/org/apache/hadoop/hbase/master/TestCatalogJanitor.java Wed Nov 30 05:43:28 2011
@@ -52,7 +52,6 @@ import org.junit.Test;
 import org.mockito.Mockito;
 
 public class TestCatalogJanitor {
-
   /**
    * Pseudo server for below tests.
    */
@@ -65,8 +64,7 @@ public class TestCatalogJanitor {
       this.c = htu.getConfiguration();
       // Set hbase.rootdir into test dir.
       FileSystem fs = FileSystem.get(this.c);
-      Path rootdir =
-        fs.makeQualified(HBaseTestingUtility.getTestDir(HConstants.HBASE_DIR));
+      Path rootdir = fs.makeQualified(new Path(this.c.get(HConstants.HBASE_DIR)));
       this.c.set(HConstants.HBASE_DIR, rootdir.toString());
       this.ct = Mockito.mock(CatalogTracker.class);
       HRegionInterface hri = Mockito.mock(HRegionInterface.class);
@@ -183,12 +181,12 @@ public class TestCatalogJanitor {
   @Test
   public void testCleanParent() throws IOException {
     HBaseTestingUtility htu = new HBaseTestingUtility();
+    setRootDirAndCleanIt(htu, "testCleanParent");
     Server server = new MockServer(htu);
     MasterServices services = new MockMasterServices(server);
     CatalogJanitor janitor = new CatalogJanitor(server, services);
     // Create regions.
-    HTableDescriptor htd = new HTableDescriptor("table");
-    htd.addFamily(new HColumnDescriptor("family"));
+    HTableDescriptor htd = createHTableDescriptor();
     HRegionInfo parent =
       new HRegionInfo(htd, Bytes.toBytes("aaa"), Bytes.toBytes("eee"));
     HRegionInfo splita =
@@ -197,12 +195,7 @@ public class TestCatalogJanitor {
       new HRegionInfo(htd, Bytes.toBytes("ccc"), Bytes.toBytes("eee"));
     // Test that when both daughter regions are in place, that we do not
     // remove the parent.
-    List<KeyValue> kvs = new ArrayList<KeyValue>();
-    kvs.add(new KeyValue(parent.getRegionName(), HConstants.CATALOG_FAMILY,
-      HConstants.SPLITA_QUALIFIER, Writables.getBytes(splita)));
-    kvs.add(new KeyValue(parent.getRegionName(), HConstants.CATALOG_FAMILY,
-      HConstants.SPLITB_QUALIFIER, Writables.getBytes(splitb)));
-    Result r = new Result(kvs);
+    Result r = createResult(parent, splita, splitb);
     // Add a reference under splitA directory so we don't clear out the parent.
     Path rootdir = services.getMasterFileSystem().getRootDir();
     Path tabledir =
@@ -220,4 +213,159 @@ public class TestCatalogJanitor {
     assertTrue(fs.delete(p, true));
     assertTrue(janitor.cleanParent(parent, r));
   }
-}
+
+  /**
+   * Make sure parent gets cleaned up even if daughter is cleaned up before it.
+   * @throws IOException
+   * @throws InterruptedException 
+   */
+  @Test
+  public void testParentCleanedEvenIfDaughterGoneFirst()
+  throws IOException, InterruptedException {
+    HBaseTestingUtility htu = new HBaseTestingUtility();
+    setRootDirAndCleanIt(htu, "testParentCleanedEvenIfDaughterGoneFirst");
+    Server server = new MockServer(htu);
+    MasterServices services = new MockMasterServices(server);
+    CatalogJanitor janitor = new CatalogJanitor(server, services);
+    final HTableDescriptor htd = createHTableDescriptor();
+
+    // Create regions: aaa->eee, aaa->ccc, aaa->bbb, bbb->ccc, etc.
+
+    // Parent
+    HRegionInfo parent = new HRegionInfo(htd.getName(), Bytes.toBytes("aaa"),
+      Bytes.toBytes("eee"));
+    // Sleep a second else the encoded name on these regions comes out
+    // same for all with same start key and made in same second.
+    Thread.sleep(1001);
+
+    // Daughter a
+    HRegionInfo splita = new HRegionInfo(htd.getName(), Bytes.toBytes("aaa"),
+      Bytes.toBytes("ccc"));
+    Thread.sleep(1001);
+    // Make daughters of daughter a; splitaa and splitab.
+    HRegionInfo splitaa = new HRegionInfo(htd.getName(), Bytes.toBytes("aaa"),
+      Bytes.toBytes("bbb"));
+    HRegionInfo splitab = new HRegionInfo(htd.getName(), Bytes.toBytes("bbb"),
+      Bytes.toBytes("ccc"));
+
+    // Daughter b
+    HRegionInfo splitb = new HRegionInfo(htd.getName(), Bytes.toBytes("ccc"),
+      Bytes.toBytes("eee"));
+    Thread.sleep(1001);
+    // Make Daughters of daughterb; splitba and splitbb.
+    HRegionInfo splitba = new HRegionInfo(htd.getName(), Bytes.toBytes("ccc"),
+      Bytes.toBytes("ddd"));
+    HRegionInfo splitbb = new HRegionInfo(htd.getName(), Bytes.toBytes("ddd"),
+      Bytes.toBytes("eee"));
+
+    // First test that our Comparator works right up in CatalogJanitor.
+    // Just fo kicks.
+    SortedMap<HRegionInfo, Result> regions =
+      new TreeMap<HRegionInfo, Result>(new CatalogJanitor.SplitParentFirstComparator());
+    // Now make sure that this regions map sorts as we expect it to.
+    regions.put(parent, createResult(parent, splita, splitb));
+    regions.put(splitb, createResult(splitb, splitba, splitbb));
+    regions.put(splita, createResult(splita, splitaa, splitab));
+    // Assert its properly sorted.
+    int index = 0;
+    for (Map.Entry<HRegionInfo, Result> e: regions.entrySet()) {
+      if (index == 0) {
+        assertTrue(e.getKey().getEncodedName().equals(parent.getEncodedName()));
+      } else if (index == 1) {
+        assertTrue(e.getKey().getEncodedName().equals(splita.getEncodedName()));
+      } else if (index == 2) {
+        assertTrue(e.getKey().getEncodedName().equals(splitb.getEncodedName()));
+      }
+      index++;
+    }
+
+    // Now play around with the cleanParent function.  Create a ref from splita
+    // up to the parent.
+    Path splitaRef =
+      createReferences(services, htd, parent, splita, Bytes.toBytes("ccc"), false);
+    // Make sure actual super parent sticks around because splita has a ref.
+    assertFalse(janitor.cleanParent(parent, regions.get(parent)));
+
+    //splitba, and split bb, do not have dirs in fs.  That means that if
+    // we test splitb, it should get cleaned up.
+    assertTrue(janitor.cleanParent(splitb, regions.get(splitb)));
+
+    // Now remove ref from splita to parent... so parent can be let go and so
+    // the daughter splita can be split (can't split if still references).
+    // BUT make the timing such that the daughter gets cleaned up before we
+    // can get a chance to let go of the parent.
+    FileSystem fs = FileSystem.get(htu.getConfiguration());
+    assertTrue(fs.delete(splitaRef, true));
+    // Create the refs from daughters of splita.
+    Path splitaaRef =
+      createReferences(services, htd, splita, splitaa, Bytes.toBytes("bbb"), false);
+    Path splitabRef =
+      createReferences(services, htd, splita, splitab, Bytes.toBytes("bbb"), true);
+
+    // Test splita.  It should stick around because references from splitab, etc.
+    assertFalse(janitor.cleanParent(splita, regions.get(splita)));
+
+    // Now clean up parent daughter first.  Remove references from its daughters.
+    assertTrue(fs.delete(splitaaRef, true));
+    assertTrue(fs.delete(splitabRef, true));
+    assertTrue(janitor.cleanParent(splita, regions.get(splita)));
+
+    // Super parent should get cleaned up now both splita and splitb are gone.
+    assertTrue(janitor.cleanParent(parent, regions.get(parent)));
+  }
+
+  private String setRootDirAndCleanIt(final HBaseTestingUtility htu,
+      final String subdir)
+  throws IOException {
+    Path testdir = HBaseTestingUtility.getTestDir(subdir);
+    FileSystem fs = FileSystem.get(htu.getConfiguration());
+    if (fs.exists(testdir)) assertTrue(fs.delete(testdir, true));
+    htu.getConfiguration().set(HConstants.HBASE_DIR, testdir.toString());
+    return htu.getConfiguration().get(HConstants.HBASE_DIR);
+  }
+
+  /**
+   * @param services Master services instance.
+   * @param htd
+   * @param parent
+   * @param daughter
+   * @param midkey
+   * @param top True if we are to write a 'top' reference.
+   * @return Path to reference we created.
+   * @throws IOException
+   */
+  private Path createReferences(final MasterServices services,
+      final HTableDescriptor htd, final HRegionInfo parent,
+      final HRegionInfo daughter, final byte [] midkey, final boolean top)
+  throws IOException {
+    Path rootdir = services.getMasterFileSystem().getRootDir();
+    Path tabledir = HTableDescriptor.getTableDir(rootdir, parent.getTableName());
+    Path storedir = Store.getStoreHomedir(tabledir, daughter.getEncodedName(),
+      htd.getColumnFamilies()[0].getName());
+    Reference ref = new Reference(midkey,
+      top? Reference.Range.top: Reference.Range.bottom);
+    long now = System.currentTimeMillis();
+    // Reference name has this format: StoreFile#REF_NAME_PARSER
+    Path p = new Path(storedir, Long.toString(now) + "." + parent.getEncodedName());
+    FileSystem fs = services.getMasterFileSystem().getFileSystem();
+    ref.write(fs, p);
+    return p;
+  }
+
+  private Result createResult(final HRegionInfo parent, final HRegionInfo a,
+      final HRegionInfo b)
+  throws IOException {
+    List<KeyValue> kvs = new ArrayList<KeyValue>();
+    kvs.add(new KeyValue(parent.getRegionName(), HConstants.CATALOG_FAMILY,
+      HConstants.SPLITA_QUALIFIER, Writables.getBytes(a)));
+    kvs.add(new KeyValue(parent.getRegionName(), HConstants.CATALOG_FAMILY,
+      HConstants.SPLITB_QUALIFIER, Writables.getBytes(b)));
+    return new Result(kvs);
+  }
+
+  private HTableDescriptor createHTableDescriptor() {
+    HTableDescriptor htd = new HTableDescriptor("t");
+    htd.addFamily(new HColumnDescriptor("f"));
+    return htd;
+  }
+}
\ No newline at end of file

Modified: hbase/branches/0.90/src/test/java/org/apache/hadoop/hbase/master/TestZKBasedOpenCloseRegion.java
URL: http://svn.apache.org/viewvc/hbase/branches/0.90/src/test/java/org/apache/hadoop/hbase/master/TestZKBasedOpenCloseRegion.java?rev=1208290&r1=1208289&r2=1208290&view=diff
==============================================================================
--- hbase/branches/0.90/src/test/java/org/apache/hadoop/hbase/master/TestZKBasedOpenCloseRegion.java (original)
+++ hbase/branches/0.90/src/test/java/org/apache/hadoop/hbase/master/TestZKBasedOpenCloseRegion.java Wed Nov 30 05:43:28 2011
@@ -32,6 +32,7 @@ import org.apache.hadoop.hbase.HBaseTest
 import org.apache.hadoop.hbase.HConstants;
 import org.apache.hadoop.hbase.HRegionInfo;
 import org.apache.hadoop.hbase.MiniHBaseCluster;
+import org.apache.hadoop.hbase.client.HBaseAdmin;
 import org.apache.hadoop.hbase.client.HTable;
 import org.apache.hadoop.hbase.client.Put;
 import org.apache.hadoop.hbase.client.Result;
@@ -111,31 +112,12 @@ public class TestZKBasedOpenCloseRegion 
     HRegionInfo hri = getNonMetaRegion(regionServer.getOnlineRegions());
     LOG.debug("Asking RS to close region " + hri.getRegionNameAsString());
 
-    AtomicBoolean closeEventProcessed = new AtomicBoolean(false);
     AtomicBoolean reopenEventProcessed = new AtomicBoolean(false);
 
-    EventHandlerListener closeListener =
-      new ReopenEventListener(hri.getRegionNameAsString(),
-          closeEventProcessed, EventType.RS_ZK_REGION_CLOSED);
-    cluster.getMaster().executorService.
-      registerListener(EventType.RS_ZK_REGION_CLOSED, closeListener);
-
-    EventHandlerListener openListener =
-      new ReopenEventListener(hri.getRegionNameAsString(),
-          reopenEventProcessed, EventType.RS_ZK_REGION_OPENED);
-    cluster.getMaster().executorService.
-      registerListener(EventType.RS_ZK_REGION_OPENED, openListener);
-
     LOG.info("Unassign " + hri.getRegionNameAsString());
-    cluster.getMaster().assignmentManager.unassign(hri);
 
-    while (!closeEventProcessed.get()) {
-      Threads.sleep(100);
-    }
+    regionServer.closeRegion(hri, false); 
 
-    while (!reopenEventProcessed.get()) {
-      Threads.sleep(100);
-    }
     
     //Test a region is reopened on a same region server.
     reopenEventProcessed.set(false);    
@@ -147,7 +129,7 @@ public class TestZKBasedOpenCloseRegion 
     assertTrue(master.isActiveMaster());
     
     hri = getNonMetaRegion(regionServer.getOnlineRegions());
-    openListener =
+    EventHandlerListener openListener =
       new ReopenEventListener(hri.getRegionNameAsString(),
           reopenEventProcessed, EventType.RS_ZK_REGION_OPENED);
     cluster.getMaster().executorService.



Mime
View raw message