From 44864c68b59a72f27279212c1316adec37e5209a Mon Sep 17 00:00:00 2001
From: Uma Maheswara Rao G <umamahesh@apache.org>
Date: Thu, 14 Aug 2014 04:20:00 +0000
Subject: [PATCH] HDFS-6783. Fix HDFS CacheReplicationMonitor rescan logic.
 Contributed by Yi Liu and Colin Patrick McCabe.

git-svn-id: https://svn.apache.org/repos/asf/hadoop/common/trunk@1617872 13f79535-47bb-0310-9956-ffa450edef68
---
 hadoop-hdfs-project/hadoop-hdfs/CHANGES.txt   |  2 +
 .../CacheReplicationMonitor.java              | 55 +++++++++++--------
 2 files changed, 35 insertions(+), 22 deletions(-)

diff --git a/hadoop-hdfs-project/hadoop-hdfs/CHANGES.txt b/hadoop-hdfs-project/hadoop-hdfs/CHANGES.txt
index 0a90370dc6..cffa1679fd 100644
--- a/hadoop-hdfs-project/hadoop-hdfs/CHANGES.txt
+++ b/hadoop-hdfs-project/hadoop-hdfs/CHANGES.txt
@@ -506,6 +506,8 @@ Release 2.6.0 - UNRELEASED
     HDFS-6247. Avoid timeouts for replaceBlock() call by sending intermediate
     responses to Balancer (vinayakumarb)
 
+    HDFS-6783. Fix HDFS CacheReplicationMonitor rescan logic. (Yi Liu via umamahesh)
+
 Release 2.5.0 - UNRELEASED
 
   INCOMPATIBLE CHANGES
diff --git a/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/blockmanagement/CacheReplicationMonitor.java b/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/blockmanagement/CacheReplicationMonitor.java
index cf5a4a6504..cfd6333469 100644
--- a/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/blockmanagement/CacheReplicationMonitor.java
+++ b/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/blockmanagement/CacheReplicationMonitor.java
@@ -103,22 +103,22 @@ public class CacheReplicationMonitor extends Thread implements Closeable {
    */
   private final Condition scanFinished;
 
-  /**
-   * Whether there are pending CacheManager operations that necessitate a
-   * CacheReplicationMonitor rescan. Protected by the CRM lock.
-   */
-  private boolean needsRescan = true;
-
-  /**
-   * Whether we are currently doing a rescan. Protected by the CRM lock.
-   */
-  private boolean isScanning = false;
-
   /**
    * The number of rescans completed. Used to wait for scans to finish.
    * Protected by the CacheReplicationMonitor lock.
    */
-  private long scanCount = 0;
+  private long completedScanCount = 0;
+
+  /**
+   * The scan we're currently performing, or -1 if no scan is in progress.
+   * Protected by the CacheReplicationMonitor lock.
+   */
+  private long curScanCount = -1;
+
+  /**
+   * The number of rescans we need to complete.  Protected by the CRM lock.
+   */
+  private long neededScanCount = 0;
 
   /**
    * True if this monitor should terminate. Protected by the CRM lock.
@@ -169,7 +169,7 @@ public void run() {
               LOG.info("Shutting down CacheReplicationMonitor");
               return;
             }
-            if (needsRescan) {
+            if (completedScanCount < neededScanCount) {
               LOG.info("Rescanning because of pending operations");
               break;
             }
@@ -182,8 +182,6 @@ public void run() {
             doRescan.await(delta, TimeUnit.MILLISECONDS);
             curTimeMs = Time.monotonicNow();
           }
-          isScanning = true;
-          needsRescan = false;
         } finally {
           lock.unlock();
         }
@@ -194,8 +192,8 @@ public void run() {
         // Update synchronization-related variables.
         lock.lock();
         try {
-          isScanning = false;
-          scanCount++;
+          completedScanCount = curScanCount;
+          curScanCount = -1;
           scanFinished.signalAll();
         } finally {
           lock.unlock();
@@ -226,16 +224,15 @@ public void waitForRescanIfNeeded() {
         "Must not hold the FSN write lock when waiting for a rescan.");
     Preconditions.checkArgument(lock.isHeldByCurrentThread(),
         "Must hold the CRM lock when waiting for a rescan.");
-    if (!needsRescan) {
+    if (neededScanCount <= completedScanCount) {
       return;
     }
     // If no scan is already ongoing, mark the CRM as dirty and kick
-    if (!isScanning) {
+    if (curScanCount < 0) {
       doRescan.signal();
     }
     // Wait until the scan finishes and the count advances
-    final long startCount = scanCount;
-    while ((!shutdown) && (startCount >= scanCount)) {
+    while ((!shutdown) && (completedScanCount < neededScanCount)) {
       try {
         scanFinished.await();
       } catch (InterruptedException e) {
@@ -253,7 +250,14 @@ public void waitForRescanIfNeeded() {
   public void setNeedsRescan() {
     Preconditions.checkArgument(lock.isHeldByCurrentThread(),
         "Must hold the CRM lock when setting the needsRescan bit.");
-    this.needsRescan = true;
+    if (curScanCount >= 0) {
+      // If there is a scan in progress, we need to wait for the scan after
+      // that.
+      neededScanCount = curScanCount + 1;
+    } else {
+      // If there is no scan in progress, we need to wait for the next scan.
+      neededScanCount = completedScanCount + 1;
+    }
   }
 
   /**
@@ -284,10 +288,17 @@ private void rescan() throws InterruptedException {
     scannedBlocks = 0;
     namesystem.writeLock();
     try {
+      lock.lock();
       if (shutdown) {
         throw new InterruptedException("CacheReplicationMonitor was " +
             "shut down.");
       }
+      curScanCount = completedScanCount + 1;
+    }
+    finally {
+      lock.unlock();
+    }
+    try {
       resetStatistics();
       rescanCacheDirectives();
       rescanCachedBlockMap();