GC collects duplicates among all repo data

backup_collector.cc

@@ -14,11 +14,11 @@ void BundleCollector::startIndex( string const & indexFn )
 
 void BundleCollector::finishIndex( string const & indexFn )
 {
+  verbosePrintf( "Chunks used: %d/%d, bundles: %d kept, %d modified, %d removed\n",
+                 indexUsedChunks, indexTotalChunks, indexKeptBundles,
+                 indexModifiedBundles, indexRemovedBundles );
   if ( indexModified )
   {
-    verbosePrintf( "Chunks used: %d/%d, bundles: %d kept, %d modified, %d removed\n",
-                   indexUsedChunks, indexTotalChunks, indexKeptBundles,
-                   indexModifiedBundles, indexRemovedBundles );
     filesToUnlink.push_back( indexFn );
     commit();
   }
@@ -75,19 +75,40 @@ void BundleCollector::finishBundle( Bundle::Id const & bundleId, BundleInfo cons
   }
   else
   {
-    if ( !deepGC )
-    {
-      chunkStorageWriter->addBundle( info, savedId );
-      dPrintf( "Keeping %s bundle\n", i.c_str() );
-      indexKeptBundles++;
-    }
-    else
+    if ( gcRepack )
     {
       filesToUnlink.push_back( Dir::addPath( bundlesPath, i ) );
       indexModified = true;
       copyUsedChunks( info );
       indexModifiedBundles++;
     }
+    else
+    {
+      if ( 0 == totalChunks )
+      {
+        if ( overallBundleSet.find ( bundleId ) == overallBundleSet.end() )
+        {
+          overallBundleSet.insert( bundleId );
+          dPrintf( "Deleting %s bundle\n", i.c_str() );
+          filesToUnlink.push_back( Dir::addPath( bundlesPath, i ) );
+          indexModified = true;
+          indexRemovedBundles++;
+        }
+        else
+        {
+          // trigger index update
+          indexModified = true;
+        }
+      }
+      else
+      {
+        if ( overallBundleSet.find ( bundleId ) == overallBundleSet.end() )
+          overallBundleSet.insert( bundleId );
+        chunkStorageWriter->addBundle( info, savedId );
+        dPrintf( "Keeping %s bundle\n", i.c_str() );
+        indexKeptBundles++;
+      }
+    }
   }
 }
 

backup_collector.hh

@@ -23,6 +23,7 @@ private:
   bool indexModified, indexNecessary;
   vector< string > filesToUnlink;
   BackupRestorer::ChunkSet overallChunkSet;
+  std::set< Bundle::Id > overallBundleSet;
 
   void copyUsedChunks( BundleInfo const & info );
 
@@ -31,7 +32,7 @@ public:
   ChunkStorage::Reader *chunkStorageReader;
   ChunkStorage::Writer *chunkStorageWriter;
   BackupRestorer::ChunkSet usedChunkSet;
-  bool deepGC;
+  bool gcRepack;
 
   void startIndex( string const & indexFn );
 

bundle.hh

@@ -42,6 +42,8 @@ struct Id
   { return memcmp( blob, other.blob, sizeof( blob ) ) == 0; }
   bool operator != ( Id const & other ) const
   { return ! operator == ( other ); }
+  bool operator < ( Id const & other ) const
+  { return memcmp( blob, other.blob, sizeof( blob ) ) < 0; }
 };
 
 STATIC_ASSERT( sizeof( Id ) == IdSize );

config.cc

@@ -104,8 +104,8 @@ void Config::prefillKeywords()
       "cache-size",
       Config::oRuntime_cacheSize,
       Config::Runtime,
-      "Cache size to use in restore process\n"
-      "Affects restore process speed directly\n"
+      "Cache size to use in restore process.\n"
+      "Affects restore process speed directly.\n"
       VALID_SUFFIXES
       "Default is %sMiB",
       Utils::numberToString( runtime.cacheSize / 1024 / 1024 )
@@ -114,25 +114,23 @@ void Config::prefillKeywords()
       "exchange",
       Config::oRuntime_exchange,
       Config::Runtime,
-      "Data to exchange between repositories in import/export process\n"
-      "Can be specified multiple times\n"
+      "Data to exchange between repositories in import/export process.\n"
+      "Can be specified multiple times.\n"
       "Valid values:\n"
       "backups - exchange backup instructions (files in backups/ directory)\n"
       "bundles - exchange bundles with data (files in bunles/ directory)\n"
       "indexes - exchange indexes of chunks (files in index/ directory)\n"
-      "No default value, you should specify it explicitly"
+      "No default value, you should specify it explicitly."
     },
 
     {
-      "gc-deep",
-      Config::oRuntime_gcDeep,
+      "gc-repack",
+      Config::oRuntime_gcRepack,
       Config::Runtime,
-      "Perform inter-bundle and inter-index deduplication\n"
-      "during garbage collection\n"
-      "You would probably need it after exchange operation\n"
-      "You could also use this switch to repack all bundles\n"
-      "Beware that this switch causes very intensive IO!\n"
-      "This switch is not used by default, specify to enable"
+      "Repack indexes and bundles during garbage collection.\n"
+      "Normally you would not need this.\n"
+      "Beware that this options causes very intensive IO!\n"
+      "Not default, you should specify it explicitly."
     },
 
     { "", Config::oBadOption, Config::None }
@@ -468,10 +466,10 @@ bool Config::parseOrValidate( const string & option, const OptionType type,
       /* NOTREACHED */
       break;
 
-    case oRuntime_gcDeep:
-      runtime.gcDeep = true;
+    case oRuntime_gcRepack:
+      runtime.gcRepack = true;
 
-      dPrintf( "runtime[gcDeep] = true\n" );
+      dPrintf( "runtime[gcRepack] = true\n" );
 
       return true;
       /* NOTREACHED */

config.hh

@@ -29,13 +29,13 @@ public:
     size_t threads;
     size_t cacheSize;
     bitset< BackupExchanger::Flags > exchange;
-    bool gcDeep;
+    bool gcRepack;
 
     // Default runtime config
     RuntimeConfig():
       threads( getNumberOfCpus() ),
       cacheSize( 40 * 1024 * 1024 ), // 40 MB
-      gcDeep ( false )
+      gcRepack ( false )
     {
     }
   };
@@ -60,7 +60,7 @@ public:
     oRuntime_threads,
     oRuntime_cacheSize,
     oRuntime_exchange,
-    oRuntime_gcDeep,
+    oRuntime_gcRepack,
 
     oDeprecated, oUnsupported
   } OpCodes;

zutils.cc

@@ -316,15 +316,13 @@ void ZCollector::gc()
 
   string fileName;
 
-  Dir::Entry entry;
-
   BundleCollector collector;
   collector.bundlesPath = getBundlesPath();
   collector.chunkStorageReader = &this->chunkStorageReader;
   collector.chunkStorageWriter = &chunkStorageWriter;
-  collector.deepGC = config.runtime.gcDeep;
+  collector.gcRepack = config.runtime.gcRepack;
 
-  verbosePrintf( "Checking used chunks...\n" );
+  verbosePrintf( "Performing garbage collection...\n" );
 
   verbosePrintf( "Searching for backups...\n" );
   vector< string > backups = BackupExchanger::findOrRebuild( getBackupsPath() );
@@ -356,6 +354,7 @@ void ZCollector::gc()
 
   string bundlesPath = getBundlesPath();
   Dir::Listing bundleLst( bundlesPath );
+  Dir::Entry entry;
   while( bundleLst.getNext( entry ) )
   {
     const string dirPath = Dir::addPath( bundlesPath, entry.getFileName());