HBASE-26707: Reduce number of renames during bulkload (#4066)

Signed-off-by: Wellington Ramos Chevreuil <wchevreuil@apache.org>

Author: BukrosSzabolcs

hbase-it/src/test/java/org/apache/hadoop/hbase/mapreduce/IntegrationTestFileBasedSFTBulkLoad.java

@@ -0,0 +1,166 @@
+/**
+ *
+ * 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.mapreduce;
+
+import org.apache.commons.lang3.RandomStringUtils;
+import org.apache.hadoop.conf.Configuration;
+import org.apache.hadoop.fs.Path;
+import org.apache.hadoop.hbase.Cell;
+import org.apache.hadoop.hbase.CellUtil;
+import org.apache.hadoop.hbase.HBaseConfiguration;
+import org.apache.hadoop.hbase.HBaseTestingUtil;
+import org.apache.hadoop.hbase.IntegrationTestBase;
+import org.apache.hadoop.hbase.IntegrationTestingUtility;
+import org.apache.hadoop.hbase.KeyValue;
+import org.apache.hadoop.hbase.TableName;
+import org.apache.hadoop.hbase.client.Admin;
+import org.apache.hadoop.hbase.client.Connection;
+import org.apache.hadoop.hbase.client.ConnectionFactory;
+import org.apache.hadoop.hbase.client.Consistency;
+import org.apache.hadoop.hbase.client.RegionLocator;
+import org.apache.hadoop.hbase.client.Result;
+import org.apache.hadoop.hbase.client.Scan;
+import org.apache.hadoop.hbase.client.TableDescriptor;
+import org.apache.hadoop.hbase.client.TableDescriptorBuilder;
+import org.apache.hadoop.hbase.coprocessor.ObserverContext;
+import org.apache.hadoop.hbase.coprocessor.RegionCoprocessor;
+import org.apache.hadoop.hbase.coprocessor.RegionCoprocessorEnvironment;
+import org.apache.hadoop.hbase.coprocessor.RegionObserver;
+import org.apache.hadoop.hbase.io.ImmutableBytesWritable;
+import org.apache.hadoop.hbase.regionserver.InternalScanner;
+import org.apache.hadoop.hbase.regionserver.storefiletracker.StoreFileTrackerFactory;
+import org.apache.hadoop.hbase.testclassification.IntegrationTests;
+import org.apache.hadoop.hbase.tool.BulkLoadHFiles;
+import org.apache.hadoop.hbase.util.Bytes;
+import org.apache.hadoop.hbase.util.EnvironmentEdgeManager;
+import org.apache.hadoop.hbase.util.RegionSplitter;
+import org.apache.hadoop.io.LongWritable;
+import org.apache.hadoop.io.NullWritable;
+import org.apache.hadoop.io.Writable;
+import org.apache.hadoop.io.WritableComparable;
+import org.apache.hadoop.io.WritableComparator;
+import org.apache.hadoop.io.WritableUtils;
+import org.apache.hadoop.mapreduce.InputFormat;
+import org.apache.hadoop.mapreduce.InputSplit;
+import org.apache.hadoop.mapreduce.Job;
+import org.apache.hadoop.mapreduce.JobContext;
+import org.apache.hadoop.mapreduce.Mapper;
+import org.apache.hadoop.mapreduce.Partitioner;
+import org.apache.hadoop.mapreduce.RecordReader;
+import org.apache.hadoop.mapreduce.Reducer;
+import org.apache.hadoop.mapreduce.TaskAttemptContext;
+import org.apache.hadoop.mapreduce.lib.output.FileOutputFormat;
+import org.apache.hadoop.util.StringUtils;
+import org.apache.hadoop.util.ToolRunner;
+import org.apache.hbase.thirdparty.com.google.common.base.Joiner;
+import org.apache.hbase.thirdparty.com.google.common.collect.Sets;
+import org.apache.hbase.thirdparty.org.apache.commons.cli.CommandLine;
+import org.junit.Test;
+import org.junit.experimental.categories.Category;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+import java.io.DataInput;
+import java.io.DataOutput;
+import java.io.IOException;
+import java.util.ArrayList;
+import java.util.List;
+import java.util.Map;
+import java.util.Optional;
+import java.util.Random;
+import java.util.Set;
+import java.util.concurrent.atomic.AtomicLong;
+import static org.junit.Assert.assertEquals;
+
+/**
+ * Test Bulk Load and MR on a distributed cluster.
+ * With FileBased StorefileTracker enabled.
+ * It starts an MR job that creates linked chains
+ *
+ * The format of rows is like this:
+ * Row Key -> Long
+ *
+ * L:<< Chain Id >> -> Row Key of the next link in the chain
+ * S:<< Chain Id >> -> The step in the chain that his link is.
+ * D:<< Chain Id >> -> Random Data.
+ *
+ * All chains start on row 0.
+ * All rk's are > 0.
+ *
+ * After creating the linked lists they are walked over using a TableMapper based Mapreduce Job.
+ *
+ * There are a few options exposed:
+ *
+ * hbase.IntegrationTestBulkLoad.chainLength
+ * The number of rows that will be part of each and every chain.
+ *
+ * hbase.IntegrationTestBulkLoad.numMaps
+ * The number of mappers that will be run.  Each mapper creates on linked list chain.
+ *
+ * hbase.IntegrationTestBulkLoad.numImportRounds
+ * How many jobs will be run to create linked lists.
+ *
+ * hbase.IntegrationTestBulkLoad.tableName
+ * The name of the table.
+ *
+ * hbase.IntegrationTestBulkLoad.replicaCount
+ * How many region replicas to configure for the table under test.
+ */
+@Category(IntegrationTests.class)
+public class IntegrationTestFileBasedSFTBulkLoad extends IntegrationTestBulkLoad {
+
+  private static final Logger LOG = LoggerFactory.getLogger(IntegrationTestFileBasedSFTBulkLoad.class);
+
+  private static String NUM_MAPS_KEY = "hbase.IntegrationTestBulkLoad.numMaps";
+  private static String NUM_IMPORT_ROUNDS_KEY = "hbase.IntegrationTestBulkLoad.numImportRounds";
+  private static String NUM_REPLICA_COUNT_KEY = "hbase.IntegrationTestBulkLoad.replicaCount";
+  private static int NUM_REPLICA_COUNT_DEFAULT = 1;
+
+  @Test
+  public void testFileBasedSFTBulkLoad() throws Exception {
+    super.testBulkLoad();
+  }
+
+  @Override
+  public void setUpCluster() throws Exception {
+    util = getTestingUtil(getConf());
+    util.getConfiguration().set(StoreFileTrackerFactory.TRACKER_IMPL,
+      "org.apache.hadoop.hbase.regionserver.storefiletracker.FileBasedStoreFileTracker");
+    util.initializeCluster(1);
+    int replicaCount = getConf().getInt(NUM_REPLICA_COUNT_KEY, NUM_REPLICA_COUNT_DEFAULT);
+    if (LOG.isDebugEnabled() && replicaCount != NUM_REPLICA_COUNT_DEFAULT) {
+      LOG.debug("Region Replicas enabled: " + replicaCount);
+    }
+
+    // Scale this up on a real cluster
+    if (util.isDistributedCluster()) {
+      util.getConfiguration().setIfUnset(NUM_MAPS_KEY,
+        Integer.toString(util.getAdmin().getRegionServers().size() * 10));
+      util.getConfiguration().setIfUnset(NUM_IMPORT_ROUNDS_KEY, "5");
+    } else {
+      util.startMiniMapReduceCluster();
+    }
+  }
+
+  public static void main(String[] args) throws Exception {
+    Configuration conf = HBaseConfiguration.create();
+    IntegrationTestingUtility.setUseDistributedCluster(conf);
+    int status =  ToolRunner.run(conf, new IntegrationTestFileBasedSFTBulkLoad(), args);
+    System.exit(status);
+  }
+}

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

@@ -7128,7 +7128,7 @@ public interface BulkLoadListener {
      * @return final path to be used for actual loading
      * @throws IOException
      */
-    String prepareBulkLoad(byte[] family, String srcPath, boolean copyFile)
+    String prepareBulkLoad(byte[] family, String srcPath, boolean copyFile, String customStaging)
         throws IOException;
 
     /**
@@ -7250,12 +7250,21 @@ public Map<byte[], List<Path>> bulkLoadHFiles(Collection<Pair<byte[], String>> f
           familyWithFinalPath.put(familyName, new ArrayList<>());
         }
         List<Pair<Path, Path>> lst = familyWithFinalPath.get(familyName);
+        String finalPath = path;
         try {
-          String finalPath = path;
+          boolean reqTmp = store.storeEngine.requireWritingToTmpDirFirst();
           if (bulkLoadListener != null) {
-            finalPath = bulkLoadListener.prepareBulkLoad(familyName, path, copyFile);
+            finalPath = bulkLoadListener.prepareBulkLoad(familyName, path, copyFile,
+              reqTmp ? null : regionDir.toString());
+          }
+          Pair<Path, Path> pair = null;
+          if (reqTmp) {
+            pair = store.preBulkLoadHFile(finalPath, seqId);
+          }
+          else {
+            Path livePath = new Path(finalPath);
+            pair = new Pair<>(livePath, livePath);
           }
-          Pair<Path, Path> pair = store.preBulkLoadHFile(finalPath, seqId);
           lst.add(pair);
         } catch (IOException ioe) {
           // A failure here can cause an atomicity violation that we currently
@@ -7265,7 +7274,7 @@ public Map<byte[], List<Path>> bulkLoadHFiles(Collection<Pair<byte[], String>> f
               " load " + Bytes.toString(p.getFirst()) + " : " + p.getSecond(), ioe);
           if (bulkLoadListener != null) {
             try {
-              bulkLoadListener.failedBulkLoad(familyName, path);
+              bulkLoadListener.failedBulkLoad(familyName, finalPath);
             } catch (Exception ex) {
               LOG.error("Error while calling failedBulkLoad for family " +
                   Bytes.toString(familyName) + " with path " + path, ex);

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

@@ -508,6 +508,10 @@ private Path preCommitStoreFile(final String familyName, final Path buildPath,
    * @throws IOException
    */
   Path commitStoreFile(final Path buildPath, Path dstPath) throws IOException {
+    // rename is not necessary in case of direct-insert stores
+    if(buildPath.equals(dstPath)){
+      return dstPath;
+    }
     // buildPath exists, therefore not doing an exists() check.
     if (!rename(buildPath, dstPath)) {
       throw new IOException("Failed rename of " + buildPath + " to " + dstPath);

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

@@ -27,6 +27,7 @@
 import java.util.Map;
 import java.util.concurrent.ConcurrentHashMap;
 import java.util.function.Consumer;
+import org.apache.commons.lang3.StringUtils;
 import org.apache.commons.lang3.mutable.MutableInt;
 import org.apache.hadoop.conf.Configuration;
 import org.apache.hadoop.fs.FileStatus;
@@ -341,27 +342,37 @@ private User getActiveUser() throws IOException {
     return user;
   }
 
-  private static class SecureBulkLoadListener implements BulkLoadListener {
+  //package-private for test purpose only
+  static class SecureBulkLoadListener implements BulkLoadListener {
     // Target filesystem
     private final FileSystem fs;
     private final String stagingDir;
     private final Configuration conf;
     // Source filesystem
     private FileSystem srcFs = null;
     private Map<String, FsPermission> origPermissions = null;
+    private Map<String, String> origSources = null;
 
     public SecureBulkLoadListener(FileSystem fs, String stagingDir, Configuration conf) {
       this.fs = fs;
       this.stagingDir = stagingDir;
       this.conf = conf;
       this.origPermissions = new HashMap<>();
+      this.origSources = new HashMap<>();
     }
 
     @Override
-    public String prepareBulkLoad(final byte[] family, final String srcPath, boolean copyFile)
-        throws IOException {
+    public String prepareBulkLoad(final byte[] family, final String srcPath, boolean copyFile,
+      String customStaging ) throws IOException {
       Path p = new Path(srcPath);
-      Path stageP = new Path(stagingDir, new Path(Bytes.toString(family), p.getName()));
+
+      //store customStaging for failedBulkLoad
+      String currentStaging = stagingDir;
+      if(StringUtils.isNotEmpty(customStaging)){
+        currentStaging = customStaging;
+      }
+
+      Path stageP = new Path(currentStaging, new Path(Bytes.toString(family), p.getName()));
 
       // In case of Replication for bulk load files, hfiles are already copied in staging directory
       if (p.equals(stageP)) {
@@ -390,11 +401,16 @@ public String prepareBulkLoad(final byte[] family, final String srcPath, boolean
         LOG.debug("Moving " + p + " to " + stageP);
         FileStatus origFileStatus = fs.getFileStatus(p);
         origPermissions.put(srcPath, origFileStatus.getPermission());
+        origSources.put(stageP.toString(), srcPath);
         if(!fs.rename(p, stageP)) {
           throw new IOException("Failed to move HFile: " + p + " to " + stageP);
         }
       }
-      fs.setPermission(stageP, PERM_ALL_ACCESS);
+
+      if(StringUtils.isNotEmpty(customStaging)) {
+        fs.setPermission(stageP, PERM_ALL_ACCESS);
+      }
+
       return stageP.toString();
     }
 
@@ -412,35 +428,37 @@ private void closeSrcFs() throws IOException {
     }
 
     @Override
-    public void failedBulkLoad(final byte[] family, final String srcPath) throws IOException {
+    public void failedBulkLoad(final byte[] family, final String stagedPath) throws IOException {
       try {
-        Path p = new Path(srcPath);
-        if (srcFs == null) {
-          srcFs = FileSystem.newInstance(p.toUri(), conf);
-        }
-        if (!FSUtils.isSameHdfs(conf, srcFs, fs)) {
-          // files are copied so no need to move them back
+        String src = origSources.get(stagedPath);
+        if(StringUtils.isEmpty(src)){
+          LOG.debug(stagedPath + " was not moved to staging. No need to move back");
           return;
         }
-        Path stageP = new Path(stagingDir, new Path(Bytes.toString(family), p.getName()));
 
-        // In case of Replication for bulk load files, hfiles are not renamed by end point during
-        // prepare stage, so no need of rename here again
-        if (p.equals(stageP)) {
-          LOG.debug(p.getName() + " is already available in source directory. Skipping rename.");
+        Path stageP = new Path(stagedPath);
+        if (!fs.exists(stageP)) {
+          throw new IOException(
+            "Missing HFile: " + stageP + ", can't be moved back to it's original place");
+        }
+
+        //we should not move back files if the original exists
+        Path srcPath = new Path(src);
+        if(srcFs.exists(srcPath)) {
+          LOG.debug(src + " is already at it's original place. No need to move.");
           return;
         }
 
-        LOG.debug("Moving " + stageP + " back to " + p);
-        if (!fs.rename(stageP, p)) {
-          throw new IOException("Failed to move HFile: " + stageP + " to " + p);
+        LOG.debug("Moving " + stageP + " back to " + srcPath);
+        if (!fs.rename(stageP, srcPath)) {
+          throw new IOException("Failed to move HFile: " + stageP + " to " + srcPath);
         }
 
         // restore original permission
-        if (origPermissions.containsKey(srcPath)) {
-          fs.setPermission(p, origPermissions.get(srcPath));
+        if (origPermissions.containsKey(stagedPath)) {
+          fs.setPermission(srcPath, origPermissions.get(src));
         } else {
-          LOG.warn("Can't find previous permission for path=" + srcPath);
+          LOG.warn("Can't find previous permission for path=" + stagedPath);
         }
       } finally {
         closeSrcFs();

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

@@ -54,6 +54,10 @@ public class TestBulkLoad extends TestBulkloadBase {
   public static final HBaseClassTestRule CLASS_RULE =
       HBaseClassTestRule.forClass(TestBulkLoad.class);
 
+  public TestBulkLoad(boolean useFileBasedSFT) {
+    super(useFileBasedSFT);
+  }
+
   @Test
   public void verifyBulkLoadEvent() throws IOException {
     TableName tableName = TableName.valueOf("test", "test");