/**
    *
    * 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.master;
  
  import java.io.FileNotFoundException;
  import java.io.IOException;
  import java.util.Comparator;
  import java.util.HashSet;
  import java.util.Map;
  import java.util.TreeMap;
  import java.util.concurrent.atomic.AtomicBoolean;
  import java.util.concurrent.atomic.AtomicInteger;
  
  import org.apache.commons.logging.Log;
  import org.apache.commons.logging.LogFactory;
  import org.apache.hadoop.classification.InterfaceAudience;
  import org.apache.hadoop.fs.FileSystem;
  import org.apache.hadoop.fs.Path;
  import org.apache.hadoop.hbase.Chore;
  import org.apache.hadoop.hbase.HColumnDescriptor;
  import org.apache.hadoop.hbase.HConstants;
  import org.apache.hadoop.hbase.HRegionInfo;
  import org.apache.hadoop.hbase.HTableDescriptor;
  import org.apache.hadoop.hbase.Server;
  import org.apache.hadoop.hbase.backup.HFileArchiver;
  import org.apache.hadoop.hbase.catalog.MetaEditor;
  import org.apache.hadoop.hbase.catalog.MetaReader;
  import org.apache.hadoop.hbase.client.MetaScanner;
  import org.apache.hadoop.hbase.client.MetaScanner.MetaScannerVisitor;
  import org.apache.hadoop.hbase.client.Result;
  import org.apache.hadoop.hbase.regionserver.HRegionFileSystem;
  import org.apache.hadoop.hbase.util.Bytes;
  import org.apache.hadoop.hbase.util.Pair;
  import org.apache.hadoop.hbase.util.PairOfSameType;
  import org.apache.hadoop.hbase.util.Triple;
  
  /**
   * A janitor for the catalog tables.  Scans the <code>.META.</code> catalog
   * table on a period looking for unused regions to garbage collect.
   */
  @InterfaceAudience.Private
  public class CatalogJanitor extends Chore {
    private static final Log LOG = LogFactory.getLog(CatalogJanitor.class.getName());
    private final Server server;
    private final MasterServices services;
    private AtomicBoolean enabled = new AtomicBoolean(true);
    private AtomicBoolean alreadyRunning = new AtomicBoolean(false);
  
    CatalogJanitor(final Server server, final MasterServices services) {
      super(server.getServerName() + "-CatalogJanitor",
        server.getConfiguration().getInt("hbase.catalogjanitor.interval", 300000),
        server);
      this.server = server;
      this.services = services;
    }
  
    @Override
    protected boolean initialChore() {
      try {
        if (this.enabled.get()) scan();
      } catch (IOException e) {
        LOG.warn("Failed initial scan of catalog table", e);
        return false;
      }
      return true;
    }
  
    /**
     * @param enabled
     */
    public boolean setEnabled(final boolean enabled) {
      return this.enabled.getAndSet(enabled);
    }
  
    boolean getEnabled() {
      return this.enabled.get();
    }
  
    @Override
    protected void chore() {
      try {
        if (this.enabled.get()) {
          scan();
       } else {
         LOG.warn("CatalogJanitor disabled! Not running scan.");
       }
     } catch (IOException e) {
       LOG.warn("Failed scan of catalog table", e);
     }
   }
 
   /**
    * Scans META and returns a number of scanned rows, and a map of merged
    * regions, and an ordered map of split parents.
    * @return triple of scanned rows, map of merged regions and map of split
    *         parent regioninfos
    * @throws IOException
    */
   Triple<Integer, Map<HRegionInfo, Result>, Map<HRegionInfo, Result>> getMergedRegionsAndSplitParents()
       throws IOException {
     return getMergedRegionsAndSplitParents(null);
   }
 
   /**
    * Scans META and returns a number of scanned rows, and a map of merged
    * regions, and an ordered map of split parents. if the given table name is
    * null, return merged regions and split parents of all tables, else only the
    * specified table
    * @param tableName null represents all tables
    * @return triple of scanned rows, and map of merged regions, and map of split
    *         parent regioninfos
    * @throws IOException
    */
   Triple<Integer, Map<HRegionInfo, Result>, Map<HRegionInfo, Result>> getMergedRegionsAndSplitParents(
       final byte[] tableName) throws IOException {
     final boolean isTableSpecified = (tableName != null && tableName.length != 0);
     // TODO: Only works with single .META. region currently.  Fix.
     final AtomicInteger count = new AtomicInteger(0);
     // Keep Map of found split parents.  There are candidates for cleanup.
     // Use a comparator that has split parents come before its daughters.
     final Map<HRegionInfo, Result> splitParents =
       new TreeMap<HRegionInfo, Result>(new SplitParentFirstComparator());
     final Map<HRegionInfo, Result> mergedRegions = new TreeMap<HRegionInfo, Result>();
     // This visitor collects split parents and counts rows in the .META. table
 
     MetaScannerVisitor visitor = new MetaScanner.MetaScannerVisitorBase() {
       @Override
       public boolean processRow(Result r) throws IOException {
         if (r == null || r.isEmpty()) return true;
         count.incrementAndGet();
         HRegionInfo info = HRegionInfo.getHRegionInfo(r);
         if (info == null) return true; // Keep scanning
         if (isTableSpecified
             && Bytes.compareTo(info.getTableName(), tableName) > 0) {
           // Another table, stop scanning
           return false;
         }
         if (info.isSplitParent()) splitParents.put(info, r);
         if (r.getValue(HConstants.CATALOG_FAMILY, HConstants.MERGEA_QUALIFIER) != null) {
           mergedRegions.put(info, r);
         }
         // Returning true means "keep scanning"
         return true;
       }
     };
 
     // Run full scan of .META. catalog table passing in our custom visitor with
     // the start row
     MetaScanner.metaScan(server.getConfiguration(), visitor, tableName);
 
     return new Triple<Integer, Map<HRegionInfo, Result>, Map<HRegionInfo, Result>>(
         count.get(), mergedRegions, splitParents);
   }
 
   /**
    * If merged region no longer holds reference to the merge regions, archive
    * merge region on hdfs and perform deleting references in .META.
    * @param mergedRegion
    * @param regionA
    * @param regionB
    * @return true if we delete references in merged region on .META. and archive
    *         the files on the file system
    * @throws IOException
    */
   boolean cleanMergeRegion(final HRegionInfo mergedRegion,
       final HRegionInfo regionA, final HRegionInfo regionB) throws IOException {
     FileSystem fs = this.services.getMasterFileSystem().getFileSystem();
     Path rootdir = this.services.getMasterFileSystem().getRootDir();
     Path tabledir = HTableDescriptor.getTableDir(rootdir,
         mergedRegion.getTableName());
     HTableDescriptor htd = getTableDescriptor(mergedRegion
         .getTableNameAsString());
     HRegionFileSystem regionFs = null;
     try {
       regionFs = HRegionFileSystem.openRegionFromFileSystem(
           this.services.getConfiguration(), fs, tabledir, mergedRegion, true);
     } catch (IOException e) {
       LOG.warn("Merged region does not exist: " + mergedRegion.getEncodedName());
     }
     if (regionFs == null || !regionFs.hasReferences(htd)) {
       LOG.debug("Deleting region " + regionA.getRegionNameAsString() + " and "
           + regionB.getRegionNameAsString()
           + " from fs because merged region no longer holds references");
       HFileArchiver.archiveRegion(this.services.getConfiguration(), fs, regionA);
       HFileArchiver.archiveRegion(this.services.getConfiguration(), fs, regionB);
       MetaEditor.deleteMergeQualifiers(server.getCatalogTracker(), mergedRegion);
       return true;
     }
     return false;
   }
 
   /**
    * Run janitorial scan of catalog <code>.META.</code> table looking for
    * garbage to collect.
    * @return number of cleaned regions
    * @throws IOException
    */
   int scan() throws IOException {
     try {
       if (!alreadyRunning.compareAndSet(false, true)) {
         return 0;
       }
       Triple<Integer, Map<HRegionInfo, Result>, Map<HRegionInfo, Result>> scanTriple =
         getMergedRegionsAndSplitParents();
       int count = scanTriple.getFirst();
       /**
        * clean merge regions first
        */
       int mergeCleaned = 0;
       Map<HRegionInfo, Result> mergedRegions = scanTriple.getSecond();
       for (Map.Entry<HRegionInfo, Result> e : mergedRegions.entrySet()) {
         HRegionInfo regionA = HRegionInfo.getHRegionInfo(e.getValue(),
             HConstants.MERGEA_QUALIFIER);
         HRegionInfo regionB = HRegionInfo.getHRegionInfo(e.getValue(),
             HConstants.MERGEB_QUALIFIER);
         if (regionA == null || regionB == null) {
           LOG.warn("Unexpected references regionA="
               + (regionA == null ? "null" : regionA.getRegionNameAsString())
               + ",regionB="
               + (regionB == null ? "null" : regionB.getRegionNameAsString())
               + " in merged region " + e.getKey().getRegionNameAsString());
         } else {
           if (cleanMergeRegion(e.getKey(), regionA, regionB)) {
             mergeCleaned++;
           }
         }
       }
       /**
        * clean split parents
        */
       Map<HRegionInfo, Result> splitParents = scanTriple.getThird();
 
       // Now work on our list of found parents. See if any we can clean up.
       int splitCleaned = 0;
       // regions whose parents are still around
       HashSet<String> parentNotCleaned = new HashSet<String>();
       for (Map.Entry<HRegionInfo, Result> e : splitParents.entrySet()) {
         if (!parentNotCleaned.contains(e.getKey().getEncodedName()) &&
             cleanParent(e.getKey(), e.getValue())) {
           splitCleaned++;
         } else {
           // We could not clean the parent, so it's daughters should not be cleaned either (HBASE-6160)
           PairOfSameType<HRegionInfo> daughters = HRegionInfo.getDaughterRegions(e.getValue());
           parentNotCleaned.add(daughters.getFirst().getEncodedName());
           parentNotCleaned.add(daughters.getSecond().getEncodedName());
         }
       }
       if ((mergeCleaned + splitCleaned) != 0) {
         LOG.info("Scanned " + count + " catalog row(s), gc'd " + mergeCleaned
             + " unreferenced merged region(s) and " + splitCleaned
             + " unreferenced parent region(s)");
       } else if (LOG.isDebugEnabled()) {
         LOG.debug("Scanned " + count + " catalog row(s), gc'd " + mergeCleaned
             + " unreferenced merged region(s) and " + splitCleaned
             + " unreferenced parent region(s)");
       }
       return mergeCleaned + splitCleaned;
     } finally {
       alreadyRunning.set(false);
     }
   }
 
   /**
    * Compare HRegionInfos in a way that has split parents sort BEFORE their
    * daughters.
    */
   static class SplitParentFirstComparator implements Comparator<HRegionInfo> {
     Comparator<byte[]> rowEndKeyComparator = new Bytes.RowEndKeyComparator();
     @Override
     public int compare(HRegionInfo left, HRegionInfo right) {
       // This comparator differs from the one HRegionInfo in that it sorts
       // parent before daughters.
       if (left == null) return -1;
       if (right == null) return 1;
       // Same table name.
       int result = Bytes.compareTo(left.getTableName(),
           right.getTableName());
       if (result != 0) return result;
       // Compare start keys.
       result = Bytes.compareTo(left.getStartKey(), right.getStartKey());
       if (result != 0) return result;
       // Compare end keys.
       result = rowEndKeyComparator.compare(left.getEndKey(), right.getEndKey());
 
       return -result; // Flip the result so parent comes first.
     }
   }
 
   /**
    * If daughters no longer hold reference to the parents, delete the parent.
    * @param parent HRegionInfo of split offlined parent
    * @param rowContent Content of <code>parent</code> row in
    * <code>metaRegionName</code>
    * @return True if we removed <code>parent</code> from meta table and from
    * the filesystem.
    * @throws IOException
    */
   boolean cleanParent(final HRegionInfo parent, Result rowContent)
   throws IOException {
     boolean result = false;
     // Check whether it is a merged region and not clean reference
     // No necessary to check MERGEB_QUALIFIER because these two qualifiers will
     // be inserted/deleted together
     if (rowContent.getValue(HConstants.CATALOG_FAMILY,
         HConstants.MERGEA_QUALIFIER) != null) {
       // wait cleaning merge region first
       return result;
     }
     // Run checks on each daughter split.
     PairOfSameType<HRegionInfo> daughters = HRegionInfo.getDaughterRegions(rowContent);
     Pair<Boolean, Boolean> a = checkDaughterInFs(parent, daughters.getFirst());
     Pair<Boolean, Boolean> b = checkDaughterInFs(parent, daughters.getSecond());
     if (hasNoReferences(a) && hasNoReferences(b)) {
       LOG.debug("Deleting region " + parent.getRegionNameAsString() +
         " because daughter splits no longer hold references");
 
       // This latter regionOffline should not be necessary but is done for now
       // until we let go of regionserver to master heartbeats.  See HBASE-3368.
       if (this.services.getAssignmentManager() != null) {
         // The mock used in testing catalogjanitor returns null for getAssignmnetManager.
         // Allow for null result out of getAssignmentManager.
         this.services.getAssignmentManager().regionOffline(parent);
       }
       FileSystem fs = this.services.getMasterFileSystem().getFileSystem();
       LOG.debug("Archiving parent region:" + parent);
       HFileArchiver.archiveRegion(this.services.getConfiguration(), fs, parent);
       MetaEditor.deleteRegion(this.server.getCatalogTracker(), parent);
       result = true;
     }
     return result;
   }
 
   /**
    * @param p A pair where the first boolean says whether or not the daughter
    * region directory exists in the filesystem and then the second boolean says
    * whether the daughter has references to the parent.
    * @return True the passed <code>p</code> signifies no references.
    */
   private boolean hasNoReferences(final Pair<Boolean, Boolean> p) {
     return !p.getFirst() || !p.getSecond();
   }
 
   /**
    * Checks if a daughter region -- either splitA or splitB -- still holds
    * references to parent.
    * @param parent Parent region
    * @param daughter Daughter region
    * @return A pair where the first boolean says whether or not the daughter
    * region directory exists in the filesystem and then the second boolean says
    * whether the daughter has references to the parent.
    * @throws IOException
    */
   Pair<Boolean, Boolean> checkDaughterInFs(final HRegionInfo parent, final HRegionInfo daughter)
   throws IOException {
     if (daughter == null)  {
       return new Pair<Boolean, Boolean>(Boolean.FALSE, Boolean.FALSE);
     }
 
     FileSystem fs = this.services.getMasterFileSystem().getFileSystem();
     Path rootdir = this.services.getMasterFileSystem().getRootDir();
     Path tabledir = HTableDescriptor.getTableDir(rootdir, daughter.getTableName());
 
     HRegionFileSystem regionFs = null;
     try {
       regionFs = HRegionFileSystem.openRegionFromFileSystem(
           this.services.getConfiguration(), fs, tabledir, daughter, true);
     } catch (IOException e) {
       LOG.warn("Daughter region does not exist: " + daughter.getEncodedName());
       return new Pair<Boolean, Boolean>(Boolean.FALSE, Boolean.FALSE);
     }
 
     boolean references = false;
     HTableDescriptor parentDescriptor = getTableDescriptor(parent.getTableNameAsString());
     for (HColumnDescriptor family: parentDescriptor.getFamilies()) {
       if ((references = regionFs.hasReferences(family.getNameAsString()))) {
         break;
       }
     }
     return new Pair<Boolean, Boolean>(Boolean.TRUE, Boolean.valueOf(references));
   }
 
   private HTableDescriptor getTableDescriptor(final String tableName)
       throws FileNotFoundException, IOException {
     return this.services.getTableDescriptors().get(tableName);
   }
 
   /**
    * Checks if the specified region has merge qualifiers, if so, try to clean
    * them
    * @param region
    * @return true if the specified region doesn't have merge qualifier now
    * @throws IOException
    */
   public boolean cleanMergeQualifier(final HRegionInfo region)
       throws IOException {
     // Get merge regions if it is a merged region and already has merge
     // qualifier
     Pair<HRegionInfo, HRegionInfo> mergeRegions = MetaReader
         .getRegionsFromMergeQualifier(this.services.getCatalogTracker(),
             region.getRegionName());
     if (mergeRegions == null
         || (mergeRegions.getFirst() == null && mergeRegions.getSecond() == null)) {
       // It doesn't have merge qualifier, no need to clean
       return true;
     }
     // It shouldn't happen, we must insert/delete these two qualifiers together
     if (mergeRegions.getFirst() == null || mergeRegions.getSecond() == null) {
       LOG.error("Merged region " + region.getRegionNameAsString()
           + " has only one merge qualifier in META.");
       return false;
     }
     return cleanMergeRegion(region, mergeRegions.getFirst(),
         mergeRegions.getSecond());
   }
 }