HBASE-5569 Do not collect deleted KVs when they are still in use by a scanner.

git-svn-id: https://svn.apache.org/repos/asf/hbase/trunk@1303220 13f79535-47bb-0310-9956-ffa450edef68

src/main/java/org/apache/hadoop/hbase/KeyValue.java

@@ -696,7 +696,7 @@ public class KeyValue implements Writable, HeapSize {
       return "empty";
     }
     return keyToString(this.bytes, this.offset + ROW_OFFSET, getKeyLength()) +
-      "/vlen=" + getValueLength();
+      "/vlen=" + getValueLength() + "/ts=" + memstoreTS;
   }
 
   /**

src/main/java/org/apache/hadoop/hbase/regionserver/ScanQueryMatcher.java

@@ -280,10 +280,15 @@ public class ScanQueryMatcher {
       if (!keepDeletedCells) {
         // first ignore delete markers if the scanner can do so, and the
         // range does not include the marker
+        //
+        // during flushes and compactions also ignore delete markers newer
+        // than the readpoint of any open scanner, this prevents deleted
+        // rows that could still be seen by a scanner from being collected
         boolean includeDeleteMarker = seePastDeleteMarkers ?
             tr.withinTimeRange(timestamp) :
             tr.withinOrAfterTimeRange(timestamp);
-        if (includeDeleteMarker) {
+        if (includeDeleteMarker
+            && kv.getMemstoreTS() <= maxReadPointToTrackVersions) {
           this.deletes.add(bytes, offset, qualLength, timestamp, type);
         }
         // Can't early out now, because DelFam come before any other keys

src/test/java/org/apache/hadoop/hbase/TestKeyValue.java

@@ -404,7 +404,7 @@ public class TestKeyValue extends TestCase {
     System.err.println("kv=" + kv);
     System.err.println("kvFromKey=" + kvFromKey);
     assertEquals(kvFromKey.toString(),
-        kv.toString().replaceAll("=[0-9]+$", "=0"));
+        kv.toString().replaceAll("=[0-9]+", "=0"));
   }
 
   @org.junit.Rule

src/test/java/org/apache/hadoop/hbase/regionserver/TestAtomicOperation.java

@@ -40,7 +40,6 @@ 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.util.Bytes;
-import org.apache.hadoop.hbase.util.EnvironmentEdgeManager;
 import org.apache.hadoop.hbase.util.EnvironmentEdgeManagerTestHelper;
 import org.junit.experimental.categories.Category;
 
@@ -55,11 +54,9 @@ public class TestAtomicOperation extends HBaseTestCase {
 
   HRegion region = null;
   private HBaseTestingUtility TEST_UTIL = new HBaseTestingUtility();
-  private final String DIR = TEST_UTIL.getDataTestDir("TestIncrement").toString();
+  private final String DIR = TEST_UTIL.getDataTestDir("TestAtomicOperation").toString();
 
 
-  private final int MAX_VERSIONS = 2;
-
   // Test names
   static final byte[] tableName = Bytes.toBytes("testtable");;
   static final byte[] qual1 = Bytes.toBytes("qual1");
@@ -260,8 +257,8 @@ public class TestAtomicOperation extends HBaseTestCase {
 
     // create 100 threads, each will alternate between adding and
     // removing a column
-    int numThreads = 100;
+    int numThreads = 10;
-    int opsPerThread = 1000;
+    int opsPerThread = 500;
     AtomicOperation[] all = new AtomicOperation[numThreads];
 
     AtomicLong timeStamps = new AtomicLong(0);
@@ -275,9 +272,11 @@ public class TestAtomicOperation extends HBaseTestCase {
           for (int i=0; i<numOps; i++) {
             try {
               // throw in some flushes
-              if (r.nextFloat() < 0.001) {
+              if (i%10==0) {
-                LOG.debug("flushing");
+                synchronized(region) {
-                region.flushcache();
+                  LOG.debug("flushing");
+                  region.flushcache();
+                }
               }
               long ts = timeStamps.incrementAndGet();
               RowMutations rm = new RowMutations(row);
@@ -342,7 +341,7 @@ public class TestAtomicOperation extends HBaseTestCase {
 
     // create 100 threads, each will alternate between adding and
     // removing a column
-    int numThreads = 100;
+    int numThreads = 10;
     int opsPerThread = 1000;
     AtomicOperation[] all = new AtomicOperation[numThreads];
 
@@ -358,9 +357,11 @@ public class TestAtomicOperation extends HBaseTestCase {
           for (int i=0; i<numOps; i++) {
             try {
               // throw in some flushes
-              if (r.nextFloat() < 0.001) {
+              if (i%10==0) {
-                LOG.debug("flushing");
+                synchronized(region) {
-                region.flushcache();
+                  LOG.debug("flushing");
+                  region.flushcache();
+                }
               }
               long ts = timeStamps.incrementAndGet();
               List<Mutation> mrm = new ArrayList<Mutation>();
@@ -386,6 +387,7 @@ public class TestAtomicOperation extends HBaseTestCase {
               RegionScanner rs = region.getScanner(s);
               List<KeyValue> r = new ArrayList<KeyValue>();
               while(rs.next(r));
+              rs.close();
               if (r.size() != 1) {
                 LOG.debug(r);
                 failures.incrementAndGet();

src/test/java/org/apache/hadoop/hbase/regionserver/TestCompaction.java

@@ -131,6 +131,7 @@ public class TestCompaction extends HBaseTestCase {
       r.delete(new Delete(results.get(0).getRow()), null, false);
       if (!result) break;
     } while(true);
+    s.close();
     // Flush
     r.flushcache();
     // Major compact.