Save progress

Signed-off-by: v01dstar <[email protected]>

src/blob_file_builder.cc

@@ -33,6 +33,8 @@ BlobFileBuilder::BlobFileBuilder(const TitanDBOptions& db_options,
     return;
 #endif
   }
+  // alignment_size_ = cf_options_.alignment_size;
+  alignment_size_ = 4 * 1024;
   WriteHeader();
 }
 
@@ -68,6 +70,7 @@ void BlobFileBuilder::Add(const BlobRecord& record,
   } else {
     encoder_.EncodeRecord(record);
     WriteEncoderData(&ctx->new_blob_index.blob_handle);
+    FillBlockWithPadding();
     out_ctx->emplace_back(std::move(ctx));
   }
 
@@ -143,15 +146,22 @@ void BlobFileBuilder::WriteEncoderData(BlobHandle* handle) {
   handle->offset = file_->GetFileSize();
   handle->size = encoder_.GetEncodedSize();
   live_data_size_ += handle->size;
+  if (alignment_size_ > 0) {
+    live_blocks_ += handle->size / alignment_size_ +
+                    (handle->size % alignment_size_ ? 1 : 0);
+  }
 
   status_ = file_->Append(encoder_.GetHeader());
   if (ok()) {
     status_ = file_->Append(encoder_.GetRecord());
     num_entries_++;
+    if (ok()) {
+      FillBlockWithPadding();
+    }
   }
 }
 
-void BlobFileBuilder::FillFSBlockWithPadding() {
+void BlobFileBuilder::FillBlockWithPadding() {
   if (alignment_size_ == 0) {
     return;
   }
@@ -211,13 +221,14 @@ Status BlobFileBuilder::Finish(OutContexts* out_ctx) {
   BlobFileFooter footer;
   // if has compression dictionary, encode it into meta blocks
   if (cf_options_.blob_file_compression_options.max_dict_bytes > 0) {
-    assert(blob_file_version_ == BlobFileHeader::kVersion2);
+    assert(blob_file_version_ >= BlobFileHeader::kVersion2);
     BlockHandle meta_index_handle;
     MetaIndexBuilder meta_index_builder;
     WriteCompressionDictBlock(&meta_index_builder);
     WriteRawBlock(meta_index_builder.Finish(), &meta_index_handle);
     footer.meta_index_handle = meta_index_handle;
   }
+  footer.alignment_size = alignment_size_;
 
   std::string buffer;
   footer.EncodeTo(&buffer);

src/blob_file_builder.h

@@ -69,7 +69,7 @@ class BlobFileBuilder {
   // caller to sync and close the file after calling Finish().
   BlobFileBuilder(const TitanDBOptions& db_options,
                   const TitanCFOptions& cf_options, WritableFileWriter* file,
-                  uint32_t blob_file_version = BlobFileHeader::kVersion2);
+                  uint32_t blob_file_version = BlobFileHeader::kVersion3);
 
   // Tries to add the record to the file
   // Notice:
@@ -123,7 +123,7 @@ class BlobFileBuilder {
   void WriteCompressionDictBlock(MetaIndexBuilder* meta_index_builder);
   void FlushSampleRecords(OutContexts* out_ctx);
   void WriteEncoderData(BlobHandle* handle);
-  void FillFSBlockWithPadding();
+  void FillBlockWithPadding();
 
   TitanCFOptions cf_options_;
   WritableFileWriter* file_;
@@ -143,6 +143,7 @@ class BlobFileBuilder {
   std::string smallest_key_;
   std::string largest_key_;
   uint64_t live_data_size_ = 0;
+  uint64_t live_blocks_ = 0;
 
   uint64_t alignment_size_ = 0;
 };

src/blob_file_iterator.cc

@@ -144,9 +144,9 @@ bool BlobFileIterator::GetBlobRecord() {
   status_ = file_->Read(IOOptions(), iterate_offset_, kRecordHeaderSize,
                         &header_buffer, header_buffer.get(),
                         nullptr /*aligned_buf*/, true /*for_compaction*/);
-  if (!status_.ok()) return;
+  if (!status_.ok()) return false;
   status_ = decoder_.DecodeHeader(&header_buffer);
-  if (!status_.ok()) return;
+  if (!status_.ok()) return false;
   // If the header buffer is all zero, it means the record is deleted (punch
   // hole).
   bool deleted = true;
@@ -174,7 +174,7 @@ bool BlobFileIterator::GetBlobRecord() {
         decoder_.DecodeRecord(&record_slice, &cur_blob_record_, &uncompressed_,
                               titan_cf_options_.memory_allocator());
   }
-  if (!status_.ok()) return;
+  if (!status_.ok()) return false;
 
   cur_record_offset_ = iterate_offset_;
   cur_record_size_ = kRecordHeaderSize + record_size;
@@ -217,6 +217,7 @@ void BlobFileIterator::PrefetchAndGet() {
     // If the record is valid (not punch-holed), we can return. Otherwise,
     // continue iterating until we find a valid record.
     if (live) return;
+    iterate_offset_ += alignment_size_;
   }
   valid_ = false;
 }

src/blob_file_manager.h

@@ -72,6 +72,12 @@ class BlobFileManager {
     (void)handles;
     return Status::OK();
   }
+
+  virtual Status BatchUpdateFiles(
+      const std::vector<std::shared_ptr<BlobFileMeta>>& files) {
+    (void)files;
+    return Status::OK();
+  }
 };
 
 }  // namespace titandb

src/blob_format.cc

@@ -141,6 +141,9 @@ void BlobFileMeta::EncodeTo(std::string* dst) const {
   PutVarint32(dst, file_level_);
   PutLengthPrefixedSlice(dst, smallest_key_);
   PutLengthPrefixedSlice(dst, largest_key_);
+  PutVarint64(dst, alignment_size_);
+  PutVarint64(dst, live_blocks_);
+  PutVarint64(dst, hole_punchable_blocks_);
 }
 
 Status BlobFileMeta::DecodeFromLegacy(Slice* src) {
@@ -171,11 +174,39 @@ Status BlobFileMeta::DecodeFrom(Slice* src) {
   return Status::OK();
 }
 
+Status BlobFileMeta::DecodeFromV3(Slice* src) {
+  if (!GetVarint64(src, &file_number_) || !GetVarint64(src, &file_size_) ||
+      !GetVarint64(src, &file_entries_) || !GetVarint32(src, &file_level_)) {
+    return Status::Corruption("BlobFileMeta decode failed");
+  }
+  Slice str;
+  if (GetLengthPrefixedSlice(src, &str)) {
+    smallest_key_.assign(str.data(), str.size());
+  } else {
+    return Status::Corruption("BlobSmallestKey Decode failed");
+  }
+  if (GetLengthPrefixedSlice(src, &str)) {
+    largest_key_.assign(str.data(), str.size());
+  } else {
+    return Status::Corruption("BlobLargestKey decode failed");
+  }
+  uint64_t alignment_size, live_blocks, hole_punchable_blocks;
+  if (!GetVarint64(src, &alignment_size) || !GetVarint64(src, &live_blocks) ||
+      !GetVarint64(src, &hole_punchable_blocks)) {
+    return Status::Corruption("BlobFileMeta decode failed");
+  }
+  alignment_size_ = alignment_size;
+  live_blocks_.store(live_blocks);
+  hole_punchable_blocks_.store(hole_punchable_blocks);
+  return Status::OK();
+}
+
 bool operator==(const BlobFileMeta& lhs, const BlobFileMeta& rhs) {
   return (lhs.file_number_ == rhs.file_number_ &&
           lhs.file_size_ == rhs.file_size_ &&
           lhs.file_entries_ == rhs.file_entries_ &&
-          lhs.file_level_ == rhs.file_level_);
+          lhs.file_level_ == rhs.file_level_ &&
+          lhs.live_blocks_.load() == rhs.live_blocks_.load());
 }
 
 void BlobFileMeta::FileStateTransit(const FileEvent& event) {
@@ -234,6 +265,10 @@ void BlobFileMeta::FileStateTransit(const FileEvent& event) {
       assert(state_ == FileState::kNormal);
       state_ = FileState::kToMerge;
       break;
+    case FileEvent::kPunchHoleOutput:
+      assert(state_ == FileState::kBeingGC);
+      state_ = FileState::kNormal;
+      break;
     case FileEvent::kReset:
       state_ = FileState::kNormal;
       break;

src/blob_format.h

@@ -38,7 +38,6 @@ namespace titandb {
 const uint64_t kBlobMaxHeaderSize = 12;
 const uint64_t kRecordHeaderSize = 9;
 const uint64_t kBlobFooterSize = 8 + BlockHandle::kMaxEncodedLength + 8 + 4;
-const std::string kAlignmentSizeBlockName = "titan.alignment_size";
 
 // Format of blob record (not fixed size):
 //
@@ -210,6 +209,7 @@ class BlobFileMeta {
     kFlushOrCompactionOutput,
     kDelete,
     kNeedMerge,
+    kPunchHoleOutput,
     kReset,  // reset file to normal for test
   };
 
@@ -229,28 +229,39 @@ class BlobFileMeta {
   BlobFileMeta(uint64_t _file_number, uint64_t _file_size,
                uint64_t _file_entries, uint32_t _file_level,
                const std::string& _smallest_key,
-               const std::string& _largest_key)
+               const std::string& _largest_key, uint64_t _alignment_size = 0,
+               uint64_t _live_blocks = 0)
       : file_number_(_file_number),
         file_size_(_file_size),
         file_entries_(_file_entries),
         file_level_(_file_level),
         smallest_key_(_smallest_key),
-        largest_key_(_largest_key) {}
+        largest_key_(_largest_key),
+        alignment_size_(_alignment_size),
+        live_blocks_(_live_blocks),
+        hole_punchable_blocks_(0) {}
 
   friend bool operator==(const BlobFileMeta& lhs, const BlobFileMeta& rhs);
 
   void EncodeTo(std::string* dst) const;
   Status DecodeFrom(Slice* src);
   Status DecodeFromLegacy(Slice* src);
+  Status DecodeFromV3(Slice* src);
+
+  void set_live_data_size(uint64_t size) { live_data_size_ = size; }
+  void set_live_blocks(uint64_t size) { live_blocks_ = size; }
 
   uint64_t file_number() const { return file_number_; }
   uint64_t file_size() const { return file_size_; }
   uint64_t live_data_size() const { return live_data_size_; }
   uint32_t file_level() const { return file_level_; }
   const std::string& smallest_key() const { return smallest_key_; }
   const std::string& largest_key() const { return largest_key_; }
+  uint64_t live_blocks() const { return live_blocks_; }
+  uint64_t hole_punchable_blocks() const { return hole_punchable_blocks_; }
+
+  uint64_t alignment_size() const { return alignment_size_; }
 
-  void set_live_data_size(int64_t size) { live_data_size_ = size; }
   uint64_t file_entries() const { return file_entries_; }
   FileState file_state() const { return state_; }
   bool is_obsolete() const { return state_ == FileState::kObsolete; }
@@ -273,6 +284,22 @@ class BlobFileMeta {
     return 1 - (static_cast<double>(live_data_size_) /
                 (file_size_ - kBlobMaxHeaderSize - kBlobFooterSize));
   }
+
+  double GetPunchHoleScore() const {
+    // Only hole-punch a file if we can at least reclaim 256 blocks and
+    // the remaining live data is more than 20% of the file size.
+    if (hole_punchable_blocks_ > 256 &&
+        double((live_blocks_ - hole_punchable_blocks_)) * 1024 * 4 /
+                file_size_ >
+            0.2) {
+      return hole_punchable_blocks_ * 1024 * 4 / file_size_;
+    }
+    return 0.0;
+  }
+
+  void set_hole_punchable_blocks(uint64_t size) {
+    hole_punchable_blocks_ = size;
+  }
   TitanInternalStats::StatsType GetDiscardableRatioLevel() const;
   void Dump(bool with_keys) const;
 
@@ -294,14 +321,18 @@ class BlobFileMeta {
   // Size of data with reference from SST files.
   //
   // Because the new generated SST is added to superversion before
-  // `OnFlushCompleted()`/`OnCompactionCompleted()` is called, so if there is a
-  // later compaction trigger by the new generated SST, the later
+  // `OnFlushCompleted()`/`OnCompactionCompleted()` is called, so if there is
+  // a later compaction trigger by the new generated SST, the later
   // `OnCompactionCompleted()` maybe called before the previous events'
   // `OnFlushCompleted()`/`OnCompactionCompleted()` is called.
   // So when state_ == kPendingLSM, it uses this to record the delta as a
   // positive number if any later compaction is trigger before previous
   // `OnCompactionCompleted()` is called.
   std::atomic<int64_t> live_data_size_{0};
+
+  uint64_t alignment_size_{0};
+  std::atomic<uint64_t> live_blocks_{0};
+  std::atomic<uint64_t> hole_punchable_blocks_{0};
   std::atomic<FileState> state_{FileState::kNone};
 };
 
@@ -321,8 +352,8 @@ class BlobFileMeta {
 //    |   Fixed32    | Fixed32 | Fixed32 |
 //    +--------------+---------+---------+
 //
-// The header is mean to be compatible with header of BlobDB blob files, except
-// we use a different magic number.
+// The header is mean to be compatible with header of BlobDB blob files,
+// except we use a different magic number.
 struct BlobFileHeader {
   // The first 32bits from $(echo titandb/blob | sha1sum).
   static const uint32_t kHeaderMagicNumber = 0x2be0a614ul;

src/blob_format_test.cc

@@ -36,7 +36,7 @@ TEST(BlobFormatTest, BlobIndex) {
 }
 
 TEST(BlobFormatTest, BlobFileMeta) {
-  BlobFileMeta input(2, 3, 0, 0, "0", "9");
+  BlobFileMeta input(2, 3, 0, 0, "0", "9", 0, 0);
   CheckCodec(input);
 }
 

src/blob_gc.cc

@@ -4,14 +4,22 @@ namespace rocksdb {
 namespace titandb {
 
 BlobGC::BlobGC(std::vector<std::shared_ptr<BlobFileMeta>>&& blob_files,
-               TitanCFOptions&& _titan_cf_options, bool need_trigger_next)
+               TitanCFOptions&& _titan_cf_options, bool need_trigger_next,
+               uint64_t cf_id, bool punch_hole)
     : inputs_(blob_files),
       titan_cf_options_(std::move(_titan_cf_options)),
-      trigger_next_(need_trigger_next) {
+      trigger_next_(need_trigger_next),
+      cf_id_(cf_id),
+      use_punch_hole_(punch_hole) {
   MarkFilesBeingGC();
 }
 
-BlobGC::~BlobGC() {}
+BlobGC::~BlobGC() {
+  // Release snapshot requires db pointer, so we can't release it internally.
+  // In case the caller forgets to release the snapshot, we assert here, prefer
+  // to crash in the runtime than leak.
+  assert(snapshot_ == nullptr);
+}
 
 void BlobGC::SetColumnFamily(ColumnFamilyHandle* cfh) { cfh_ = cfh; }
 
@@ -40,5 +48,12 @@ void BlobGC::ReleaseGcFiles() {
   }
 }
 
+void BlobGC::ReleaseSnapshot(DB* db) {
+  if (snapshot_ != nullptr) {
+    db->ReleaseSnapshot(snapshot_);
+    snapshot_ = nullptr;
+  }
+}
+
 }  // namespace titandb
 }  // namespace rocksdb