incorporating review comments

hbase-common/src/main/java/org/apache/hadoop/hbase/HConstants.java

@@ -1473,7 +1473,7 @@ public enum OperationStatusCode {
 
   // Regions Recovery based on high storeFileRefCount threshold value
   public static final String STORE_FILE_REF_COUNT_THRESHOLD =
-    "hbase.regions.recovery.store.file.count";
+    "hbase.regions.recovery.store.file.ref.count";
 
   // default -1 indicates there is no threshold on high storeRefCount
   public static final int DEFAULT_STORE_FILE_REF_COUNT_THRESHOLD = -1;

hbase-common/src/main/resources/hbase-default.xml

@@ -1912,15 +1912,22 @@ possible configurations would overwhelm and obscure the important.
     </description>
   </property>
   <property>
-    <name>hbase.regions.recovery.store.file.count</name>
+    <name>hbase.regions.recovery.store.file.ref.count</name>
     <value>-1</value>
     <description>
-      Store files Ref Count threshold value considered
-      for reopening regions. Any region with store files
-      ref count > this value would be eligible for
-      reopening by master. Default value -1 indicates
-      this feature is turned off. Only positive integer
-      value should be provided to enable the feature.
+      Very large ref count on a file indicates
+      that it is a ref leak on that object. Such files
+      can not be removed even after it is invalidated
+      via compaction. Only way to recover in such
+      scenario is to reopen the region which can
+      release all resources, like the refcount, leases, etc.
+      This config represents Store files Ref Count threshold
+      value considered for reopening regions.
+      Any region with store files ref count > this value
+      would be eligible for reopening by master.
+      Default value -1 indicates this feature is turned off.
+      Only positive integer value should be provided to enable
+      this feature.
     </description>
   </property>
 </configuration>

hbase-server/src/main/java/org/apache/hadoop/hbase/master/HMaster.java

@@ -1471,7 +1471,7 @@ private void startServiceThreads() throws IOException {
     getChoreService().scheduleChore(hfileCleaner);
 
     // Regions Reopen based on very high storeFileRefCount is considered enabled
-    // only if hbase.regions.recovery.store.file.count has value > 0
+    // only if hbase.regions.recovery.store.file.ref.count has value > 0
     final int maxStoreFileRefCount = conf.getInt(
       HConstants.STORE_FILE_REF_COUNT_THRESHOLD,
       HConstants.DEFAULT_STORE_FILE_REF_COUNT_THRESHOLD);

hbase-server/src/main/java/org/apache/hadoop/hbase/master/RegionsRecoveryChore.java

@@ -141,9 +141,10 @@ private Map<TableName, List<byte[]>> getTableToRegionsByRefCount(
         // Here, we take max ref count of all store files and not the cumulative
         // count of all store files
         final int maxStoreFileRefCount = regionMetrics.getMaxStoreFileRefCount();
-        // ignore store file ref count threshold <= 0 (default is -1 i.e. disabled)
-        if (storeFileRefCountThreshold > 0 && maxStoreFileRefCount > storeFileRefCountThreshold) {
-          prepareTableToReopenRegionsMap(tableToReopenRegionsMap, regionMetrics,
+
+        if (maxStoreFileRefCount > storeFileRefCountThreshold) {
+          final byte[] regionName = regionMetrics.getRegionName();
+          prepareTableToReopenRegionsMap(tableToReopenRegionsMap, regionName,
             maxStoreFileRefCount);
         }
       }
@@ -154,9 +155,8 @@ private Map<TableName, List<byte[]>> getTableToRegionsByRefCount(
 
   private void prepareTableToReopenRegionsMap(
       final Map<TableName, List<byte[]>> tableToReopenRegionsMap,
-      final RegionMetrics regionMetrics, final int regionStoreRefCount) {
+      final byte[] regionName, final int regionStoreRefCount) {
 
-    final byte[] regionName = regionMetrics.getRegionName();
     final RegionInfo regionInfo = hMaster.getAssignmentManager().getRegionInfo(regionName);
     final TableName tableName = regionInfo.getTable();
     if (TableName.isMetaTableName(tableName)) {
@@ -166,9 +166,7 @@ private void prepareTableToReopenRegionsMap(
     }
     LOG.warn("Region {} for Table {} has high storeFileRefCount {}, considering it for reopen..",
       regionInfo.getRegionNameAsString(), tableName, regionStoreRefCount);
-    if (!tableToReopenRegionsMap.containsKey(tableName)) {
-      tableToReopenRegionsMap.put(tableName, new ArrayList<>());
-    }
+    tableToReopenRegionsMap.putIfAbsent(tableName, new ArrayList<>());
     tableToReopenRegionsMap.get(tableName).add(regionName);
 
   }

hbase-server/src/main/java/org/apache/hadoop/hbase/master/procedure/ReopenTableRegionsProcedure.java

@@ -57,25 +57,25 @@ public class ReopenTableRegionsProcedure
 
   // Specify specific regions of a table to reopen.
   // if specified null, all regions of the table will be reopened.
-  private final List<byte[]> regionNamesList;
+  private final List<byte[]> regionNames;
 
   private List<HRegionLocation> regions = Collections.emptyList();
 
   private RetryCounter retryCounter;
 
   public ReopenTableRegionsProcedure() {
-    regionNamesList = null;
+    regionNames = null;
   }
 
   public ReopenTableRegionsProcedure(TableName tableName) {
     this.tableName = tableName;
-    this.regionNamesList = null;
+    this.regionNames = null;
   }
 
   public ReopenTableRegionsProcedure(final TableName tableName,
-      final List<byte[]> regionNamesList) {
+      final List<byte[]> regionNames) {
     this.tableName = tableName;
-    this.regionNamesList = regionNamesList;
+    this.regionNames = regionNames;
   }
 
   @Override
@@ -109,9 +109,9 @@ protected Flow executeFromState(MasterProcedureEnv env, ReopenTableRegionsState
           LOG.info("Table {} is disabled, give up reopening its regions", tableName);
           return Flow.NO_MORE_STATE;
         }
-        List<HRegionLocation> tableRegionsForReopen = env.getAssignmentManager()
+        List<HRegionLocation> tableRegions = env.getAssignmentManager()
           .getRegionStates().getRegionsOfTableForReopen(tableName);
-        regions = prepareRegionsForReopen(tableRegionsForReopen);
+        regions = getRegionLocationsForReopen(tableRegions);
         setNextState(ReopenTableRegionsState.REOPEN_TABLE_REGIONS_REOPEN_REGIONS);
         return Flow.HAS_MORE_STATE;
       case REOPEN_TABLE_REGIONS_REOPEN_REGIONS:
@@ -167,13 +167,13 @@ protected Flow executeFromState(MasterProcedureEnv env, ReopenTableRegionsState
     }
   }
 
-  private List<HRegionLocation> prepareRegionsForReopen(
+  private List<HRegionLocation> getRegionLocationsForReopen(
       List<HRegionLocation> tableRegionsForReopen) {
 
     List<HRegionLocation> regionsToReopen = new ArrayList<>();
-    if (CollectionUtils.isNotEmpty(regionNamesList) &&
+    if (CollectionUtils.isNotEmpty(regionNames) &&
       CollectionUtils.isNotEmpty(tableRegionsForReopen)) {
-      for (byte[] regionName : regionNamesList) {
+      for (byte[] regionName : regionNames) {
         for (HRegionLocation hRegionLocation : tableRegionsForReopen) {
           if (Bytes.equals(regionName, hRegionLocation.getRegion().getRegionName())) {
             regionsToReopen.add(hRegionLocation);

hbase-server/src/test/java/org/apache/hadoop/hbase/master/TestRegionsRecoveryChore.java

@@ -129,7 +129,7 @@ public void testRegionReopensWithStoreRefConfig() throws Exception {
     }
     Stoppable stoppable = new StoppableImplementation();
     Configuration configuration = getCustomConf();
-    configuration.setInt("hbase.regions.recovery.store.file.count", 300);
+    configuration.setInt("hbase.regions.recovery.store.file.ref.count", 300);
     regionsRecoveryChore = new RegionsRecoveryChore(stoppable, configuration, hMaster);
     regionsRecoveryChore.chore();
 
@@ -144,6 +144,43 @@ public void testRegionReopensWithStoreRefConfig() throws Exception {
       .getRegionInfo(Mockito.any());
   }
 
+  @Test
+  public void testRegionReopensWithLessThreshold() throws Exception {
+    regionNo = 0;
+    ClusterMetrics clusterMetrics = TestRegionsRecoveryChore.getClusterMetrics(4);
+    final Map<ServerName, ServerMetrics> serverMetricsMap =
+      clusterMetrics.getLiveServerMetrics();
+    LOG.debug("All Region Names with refCount....");
+    for (ServerMetrics serverMetrics : serverMetricsMap.values()) {
+      Map<byte[], RegionMetrics> regionMetricsMap = serverMetrics.getRegionMetrics();
+      for (RegionMetrics regionMetrics : regionMetricsMap.values()) {
+        LOG.debug("name: " + new String(regionMetrics.getRegionName()) + " refCount: " +
+          regionMetrics.getStoreRefCount());
+      }
+    }
+    Mockito.when(hMaster.getClusterMetrics()).thenReturn(clusterMetrics);
+    Mockito.when(hMaster.getAssignmentManager()).thenReturn(assignmentManager);
+    for (byte[] regionName : REGION_NAME_LIST) {
+      Mockito.when(assignmentManager.getRegionInfo(regionName))
+        .thenReturn(TestRegionsRecoveryChore.getRegionInfo(regionName));
+    }
+    Stoppable stoppable = new StoppableImplementation();
+    Configuration configuration = getCustomConf();
+    configuration.setInt("hbase.regions.recovery.store.file.ref.count", 400);
+    regionsRecoveryChore = new RegionsRecoveryChore(stoppable, configuration, hMaster);
+    regionsRecoveryChore.chore();
+
+    // Verify that we need to reopen regions of only 1 table
+    Mockito.verify(hMaster, Mockito.times(1)).reopenRegions(Mockito.any(), Mockito.anyList(),
+      Mockito.anyLong(), Mockito.anyLong());
+    Mockito.verify(hMaster, Mockito.times(1)).getClusterMetrics();
+
+    // Verify that we need to reopen only 1 region with refCount > 400
+    Mockito.verify(hMaster, Mockito.times(1)).getAssignmentManager();
+    Mockito.verify(assignmentManager, Mockito.times(1))
+      .getRegionInfo(Mockito.any());
+  }
+
   @Test
   public void testRegionReopensWithoutStoreRefConfig() throws Exception {
     regionNo = 0;
@@ -166,7 +203,7 @@ public void testRegionReopensWithoutStoreRefConfig() throws Exception {
     }
     Stoppable stoppable = new StoppableImplementation();
     Configuration configuration = getCustomConf();
-    configuration.unset("hbase.regions.recovery.store.file.count");
+    configuration.unset("hbase.regions.recovery.store.file.ref.count");
     regionsRecoveryChore = new RegionsRecoveryChore(stoppable, configuration, hMaster);
     regionsRecoveryChore.chore();
 

src/main/asciidoc/_chapters/hbase-default.adoc

@@ -2178,17 +2178,24 @@ The percent of region server RPC threads failed to abort RS.
 `1200000`
 
 
-[[hbase.regions.recovery.store.file.count]]
-*`hbase.regions.recovery.store.file.count`*::
-+
-.Description
-
-      Store files Ref Count threshold value considered
-      for reopening regions. Any region with store files
-      ref count > this value would be eligible for
-      reopening by master. Default value -1 indicates
-      this feature is turned off. Only positive integer
-      value should be provided to enable the feature.
+[[hbase.regions.recovery.store.file.ref.count]]
+*`hbase.regions.recovery.store.file.ref.count`*::
++
+.Description
+
+      Very large ref count on a file indicates
+      that it is a ref leak on that object. Such files
+      can not be removed even after it is invalidated
+      via compaction. Only way to recover in such
+      scenario is to reopen the region which can
+      release all resources, like the refcount, leases, etc.
+      This config represents Store files Ref Count threshold
+      value considered for reopening regions.
+      Any region with store files ref count > this value
+      would be eligible for reopening by master.
+      Default value -1 indicates this feature is turned off.
+      Only positive integer value should be provided to enable
+      this feature.
 
 +
 .Default