HBASE-26242 Allow split when store file count larger than the configed blocking file count (#3652)

Signed-off-by: Andrew Purtell <apurtell@apache.org>

hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/CompactSplit.java

@@ -200,7 +200,9 @@ public class CompactSplit implements CompactionRequester, PropagatingConfigurati
   }
 
   public synchronized boolean requestSplit(final Region r) {
-    // don't split regions that are blocking
+    // Don't split regions that are blocking is the default behavior.
+    // But in some circumstances, split here is needed to prevent the region size from
+    // continuously growing, as well as the number of store files, see HBASE-26242.
     HRegion hr = (HRegion)r;
     try {
       if (shouldSplitRegion() && hr.getCompactPriority() >= PRIORITY_USER) {
@@ -218,14 +220,14 @@ public class CompactSplit implements CompactionRequester, PropagatingConfigurati
     return false;
   }
 
-  public synchronized void requestSplit(final Region r, byte[] midKey) {
+  private synchronized void requestSplit(final Region r, byte[] midKey) {
     requestSplit(r, midKey, null);
   }
 
   /*
    * The User parameter allows the split thread to assume the correct user identity
    */
-  public synchronized void requestSplit(final Region r, byte[] midKey, User user) {
+  private synchronized void requestSplit(final Region r, byte[] midKey, User user) {
     if (midKey == null) {
       LOG.debug("Region " + r.getRegionInfo().getRegionNameAsString() +
         " not splittable because midkey=null");
@@ -487,9 +489,9 @@ public class CompactSplit implements CompactionRequester, PropagatingConfigurati
   }
 
   private boolean shouldSplitRegion() {
-    if(server.getNumberOfOnlineRegions() > 0.9*regionSplitLimit) {
+    if (server.getNumberOfOnlineRegions() > 0.9 * regionSplitLimit) {
       LOG.warn("Total number of regions is approaching the upper limit " + regionSplitLimit + ". "
-          + "Please consider taking a look at http://hbase.apache.org/book.html#ops.regionmgt");
+        + "Please consider taking a look at http://hbase.apache.org/book.html#ops.regionmgt");
     }
     return (regionSplitLimit > server.getNumberOfOnlineRegions());
   }
@@ -657,11 +659,14 @@ public class CompactSplit implements CompactionRequester, PropagatingConfigurati
               this + "; duration=" + StringUtils.formatTimeDiff(now, start));
         if (completed) {
           // degenerate case: blocked regions require recursive enqueues
-          if (store.getCompactPriority() <= 0) {
+          if (region.getCompactPriority() < Store.PRIORITY_USER
+            && store.getCompactPriority() <= 0) {
             requestSystemCompaction(region, store, "Recursive enqueue");
           } else {
             // see if the compaction has caused us to exceed max region size
-            requestSplit(region);
+            if (!requestSplit(region) && store.getCompactPriority() <= 0) {
+              requestSystemCompaction(region, store, "Recursive enqueue");
+            }
           }
         }
       } catch (IOException ex) {

hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/HRegion.java

@@ -267,6 +267,10 @@ public class HRegion implements HeapSize, PropagatingConfigurationObserver, Regi
   public static final String COMPACTION_AFTER_BULKLOAD_ENABLE =
       "hbase.compaction.after.bulkload.enable";
 
+  /** Config for allow split when file count greater than the configured blocking file count*/
+  public static final String SPLIT_IGNORE_BLOCKING_ENABLED_KEY =
+      "hbase.hregion.split.ignore.blocking.enabled";
+
   /**
    * This is for for using HRegion as a local storage, where we may put the recovered edits in a
    * special place. Once this is set, we will only replay the recovered edits under this directory
@@ -8196,6 +8200,10 @@ public class HRegion implements HeapSize, PropagatingConfigurationObserver, Regi
    * @return The priority that this region should have in the compaction queue
    */
   public int getCompactPriority() {
+    if (checkSplit().isPresent() && conf.getBoolean(SPLIT_IGNORE_BLOCKING_ENABLED_KEY, false)) {
+      // if a region should split, split it before compact
+      return Store.PRIORITY_USER;
+    }
     return stores.values().stream().mapToInt(HStore::getCompactPriority).min()
         .orElse(Store.NO_PRIORITY);
   }

hbase-server/src/test/java/org/apache/hadoop/hbase/regionserver/TestSplitWithBlockingFiles.java

@@ -0,0 +1,149 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+package org.apache.hadoop.hbase.regionserver;
+
+import static org.apache.hadoop.hbase.regionserver.HRegion.SPLIT_IGNORE_BLOCKING_ENABLED_KEY;
+import static org.apache.hadoop.hbase.regionserver.Store.PRIORITY_USER;
+import static org.junit.Assert.assertEquals;
+import static org.junit.Assert.assertFalse;
+import static org.junit.Assert.assertNotNull;
+import static org.junit.Assert.assertTrue;
+import java.util.List;
+import org.apache.hadoop.hbase.HBaseClassTestRule;
+import org.apache.hadoop.hbase.HBaseTestingUtil;
+import org.apache.hadoop.hbase.HConstants;
+import org.apache.hadoop.hbase.TableName;
+import org.apache.hadoop.hbase.client.Admin;
+import org.apache.hadoop.hbase.client.ColumnFamilyDescriptorBuilder;
+import org.apache.hadoop.hbase.client.Put;
+import org.apache.hadoop.hbase.client.ResultScanner;
+import org.apache.hadoop.hbase.client.Scan;
+import org.apache.hadoop.hbase.client.Table;
+import org.apache.hadoop.hbase.client.TableDescriptor;
+import org.apache.hadoop.hbase.client.TableDescriptorBuilder;
+import org.apache.hadoop.hbase.master.assignment.SplitTableRegionProcedure;
+import org.apache.hadoop.hbase.master.procedure.MasterProcedureEnv;
+import org.apache.hadoop.hbase.procedure2.ProcedureExecutor;
+import org.apache.hadoop.hbase.procedure2.ProcedureTestingUtility;
+import org.apache.hadoop.hbase.testclassification.MediumTests;
+import org.apache.hadoop.hbase.util.Bytes;
+import org.junit.AfterClass;
+import org.junit.Assert;
+import org.junit.BeforeClass;
+import org.junit.ClassRule;
+import org.junit.Test;
+import org.junit.experimental.categories.Category;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+import org.apache.hbase.thirdparty.com.google.common.io.Closeables;
+
+@Category({ MediumTests.class})
+public class TestSplitWithBlockingFiles {
+
+  @ClassRule
+  public static final HBaseClassTestRule CLASS_RULE =
+    HBaseClassTestRule.forClass(TestSplitWithBlockingFiles.class);
+
+  private static final Logger LOG = LoggerFactory.getLogger(TestSplitWithBlockingFiles.class);
+
+  protected static final HBaseTestingUtil UTIL = new HBaseTestingUtil();
+  private static TableName TABLE_NAME = TableName.valueOf("test");
+  private static Admin ADMIN;
+  private static byte[] CF = Bytes.toBytes("cf");
+  private static Table TABLE;
+
+
+  @BeforeClass
+  public static void setupCluster() throws Exception {
+    UTIL.getConfiguration().setLong(HConstants.HREGION_MAX_FILESIZE, 8 * 2 * 10240L);
+    UTIL.getConfiguration().setInt(HStore.BLOCKING_STOREFILES_KEY, 1);
+    UTIL.getConfiguration().set(HConstants.HBASE_REGION_SPLIT_POLICY_KEY,
+      ConstantSizeRegionSplitPolicy.class.getName());
+    UTIL.getConfiguration().setBoolean(SPLIT_IGNORE_BLOCKING_ENABLED_KEY, true);
+    UTIL.startMiniCluster(1);
+    ADMIN = UTIL.getAdmin();
+    TableDescriptor td = TableDescriptorBuilder.newBuilder(TABLE_NAME)
+      .setColumnFamily(
+        ColumnFamilyDescriptorBuilder.newBuilder(CF).setBlocksize(1000).build()).build();
+    TABLE = UTIL.createTable(td, null);
+    UTIL.waitTableAvailable(TABLE_NAME);
+  }
+
+  @AfterClass
+  public static void cleanupTest() throws Exception {
+    Closeables.close(TABLE, true);
+    UTIL.shutdownMiniCluster();
+  }
+
+  @Test
+  public void testSplitIgnoreBlockingFiles() throws Exception {
+    ADMIN.splitSwitch(false, true);
+    byte[] value = new byte[1024];
+    for (int m = 0; m < 10; m++) {
+      String rowPrefix = "row" + m;
+      for (int i = 0; i < 10; i++) {
+        Put p = new Put(Bytes.toBytes(rowPrefix + i));
+        p.addColumn(CF, Bytes.toBytes("qualifier"), value);
+        p.addColumn(CF, Bytes.toBytes("qualifier2"), value);
+        TABLE.put(p);
+      }
+      ADMIN.flush(TABLE_NAME);
+    }
+    Scan scan = new Scan();
+    ResultScanner results = TABLE.getScanner(scan);
+    int count = 0;
+    while (results.next() != null) {
+      count++;
+    }
+    Assert.assertEquals("There should be 100 rows!", 100, count);
+    List<HRegion> regions = UTIL.getMiniHBaseCluster().getRegionServer(0).getRegions();
+    regions.removeIf(r -> !r.getRegionInfo().getTable().equals(TABLE_NAME));
+    assertEquals(1, regions.size());
+    assertNotNull(regions.get(0).getSplitPolicy().getSplitPoint());
+    assertTrue(regions.get(0).getCompactPriority() >= PRIORITY_USER);
+    assertTrue(UTIL.getMiniHBaseCluster().getRegionServer(0).getCompactSplitThread()
+      .requestSplit(regions.get(0)));
+
+    // split region
+    ADMIN.splitSwitch(true, true);
+    MasterProcedureEnv env =
+      UTIL.getMiniHBaseCluster().getMaster().getMasterProcedureExecutor().getEnvironment();
+    final ProcedureExecutor<MasterProcedureEnv> executor =
+      UTIL.getMiniHBaseCluster().getMaster().getMasterProcedureExecutor();
+    SplitTableRegionProcedure splitProcedure =
+      new SplitTableRegionProcedure(env, regions.get(0).getRegionInfo(), Bytes.toBytes("row5"));
+    executor.submitProcedure(splitProcedure);
+    ProcedureTestingUtility.waitProcedure(executor, splitProcedure.getProcId());
+
+    regions = UTIL.getMiniHBaseCluster().getRegionServer(0).getRegions();
+    regions.removeIf(r -> !r.getRegionInfo().getTable().equals(TABLE_NAME));
+    assertEquals(2, regions.size());
+    scan = new Scan();
+    results = TABLE.getScanner(scan);
+    count = 0;
+    while (results.next() != null) {
+      count++;
+    }
+    Assert.assertEquals("There should be 100 rows!", 100, count);
+    for (HRegion region : regions) {
+      assertTrue(region.getCompactPriority() < PRIORITY_USER);
+      assertFalse(
+        UTIL.getMiniHBaseCluster().getRegionServer(0).getCompactSplitThread().requestSplit(region));
+    }
+  }
+}