diff --git a/CMakeLists.txt b/CMakeLists.txt
index 974f190..a3a5eb6 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -117,7 +117,9 @@ endif(BUILD_SHARED_LIBS)
 # Must be included before CMAKE_INSTALL_INCLUDEDIR is used.
 include(GNUInstallDirs)
 
-add_library(leveldb)
+add_library(leveldb
+        db/vlog_reader.h
+        db/vlog_reader.cpp)
 target_sources(leveldb
   PRIVATE
     "${PROJECT_BINARY_DIR}/${LEVELDB_PORT_CONFIG_DIR}/port_config.h"
diff --git a/db/db_impl.cc b/db/db_impl.cc
index 3a5c225..eacde23 100644
--- a/db/db_impl.cc
+++ b/db/db_impl.cc
@@ -36,8 +36,12 @@
 #include "util/logging.h"
 #include "util/mutexlock.h"
 
+#include "db/vlog_reader.h"
+
 namespace leveldb {
 
+using namespace log;
+
 const int kNumNonTableCacheFiles = 10;
 
 // Information kept for every waiting writer
@@ -1118,6 +1122,25 @@ int64_t DBImpl::TEST_MaxNextLevelOverlappingBytes() {
   return versions_->MaxNextLevelOverlappingBytes();
 }
 
+bool DBImpl::ParseVlogValue(Slice key_value, Slice key,
+                            std::string& value, uint64_t val_size) {
+  Slice k_v = key_value;
+  if (k_v[0] != kTypeSeparation) return false;
+  k_v.remove_prefix(1);
+
+  Slice vlog_key;
+  Slice vlog_value;
+  if (GetLengthPrefixedSlice(&k_v, &vlog_key)
+      && vlog_key == key
+      && GetLengthPrefixedSlice(&k_v, &vlog_value)
+      && vlog_value.size() == val_size) {
+    value = vlog_value.ToString();
+    return true;
+  } else {
+    return false;
+  }
+}
+
 Status DBImpl::Get(const ReadOptions& options, const Slice& key,
                    std::string* value) {
   Status s;
@@ -1162,6 +1185,53 @@ Status DBImpl::Get(const ReadOptions& options, const Slice& key,
   mem->Unref();
   if (imm != nullptr) imm->Unref();
   current->Unref();
+
+  /* Vlog 读取 value */
+  if (s.ok() && s.IsSeparated()) {
+
+    struct VlogReporter : public VlogReader::Reporter {
+      Status* status;
+      void Corruption(size_t bytes, const Status& s) override {
+        if (this->status->ok()) *this->status = s;
+      }
+    };
+
+    VlogReporter reporter;
+    Slice vlog_ptr(*value);
+    uint64_t file_no;
+    uint64_t offset;
+    uint64_t val_size;
+    size_t key_size = key.size();
+
+    GetVarint64(&vlog_ptr, &file_no);
+    GetVarint64(&vlog_ptr, &offset);
+    GetVarint64(&vlog_ptr, &val_size);
+    uint64_t encoded_len = 1 + VarintLength(key_size) + key.size() + VarintLength(val_size) + val_size;
+
+    std::string fname = LogFileName(dbname_, file_no);
+    RandomAccessFile* file;
+    s = env_->NewRandomAccessFile(fname,&file);
+    if (!s.ok()) {
+      return s;
+    }
+
+    VlogReader vlogReader(file, &reporter);
+    Slice key_value;
+    Slice ret_value;
+    char* scratch = new char[encoded_len];
+
+    if (vlogReader.ReadValue(offset, encoded_len, &key_value, scratch)) {
+      value->clear();
+      if (!ParseVlogValue(key_value, key, *value, val_size)) {
+        s = Status::Corruption("value in vlog isn't match with given key");
+      }
+    } else {
+      s = Status::Corruption("read vlog error");
+    }
+
+    delete file;
+  }
+
   return s;
 }
 
diff --git a/db/db_impl.h b/db/db_impl.h
index 91f4fd4..27c09d2 100644
--- a/db/db_impl.h
+++ b/db/db_impl.h
@@ -53,6 +53,8 @@ class DBImpl : public DB {
   void GetApproximateSizes(const Range* range, int n, uint64_t* sizes) override;
   void CompactRange(const Slice* begin, const Slice* end) override;
 
+  bool ParseVlogValue(Slice key_value, Slice key, std::string& value, uint64_t val_size);
+
   // Extra methods (for testing) that are not in the public DB interface
 
   // Compact any files in the named level that overlap [*begin,*end]
diff --git a/db/dbformat.h b/db/dbformat.h
index a1c30ed..e74af21 100644
--- a/db/dbformat.h
+++ b/db/dbformat.h
@@ -51,7 +51,7 @@ class InternalKey;
 // Value types encoded as the last component of internal keys.
 // DO NOT CHANGE THESE ENUM VALUES: they are embedded in the on-disk
 // data structures.
-enum ValueType { kTypeDeletion = 0x0, kTypeValue = 0x1 };
+enum ValueType { kTypeDeletion = 0x0, kTypeValue = 0x1, kTypeSeparation = 0x2};
 // kValueTypeForSeek defines the ValueType that should be passed when
 // constructing a ParsedInternalKey object for seeking to a particular
 // sequence number (since we sort sequence numbers in decreasing order
diff --git a/db/fields.cpp b/db/fields.cpp
index 9fb5357..fb845eb 100644
--- a/db/fields.cpp
+++ b/db/fields.cpp
@@ -171,7 +171,7 @@ std::string& Fields::operator[](const std::string& field_name) {
 }
 
 /* 通过若干个字段查询 Key */
- std::vector<std::string> Fields::FindKeysByFields(leveldb::DB* db, const FieldArray& fields) {
+std::vector<std::string> Fields::FindKeysByFields(leveldb::DB* db, const FieldArray& fields) {
     Fields to_fields = Fields(fields);
     to_fields.Fields::SortFields();
     FieldArray search_fields_ = to_fields.fields_;
@@ -199,7 +199,7 @@ std::string& Fields::operator[](const std::string& field_name) {
     delete it;
 
     return find_keys;
- }
+}
 
 //std::vector<std::string> Fields::FindKeysByFields(leveldb::DB* db, const FieldArray& fields) {
 //    Fields to_fields = Fields(fields);
diff --git a/db/log_format.h b/db/log_format.h
index 356e69f..1ea5429 100644
--- a/db/log_format.h
+++ b/db/log_format.h
@@ -29,6 +29,16 @@ static const int kBlockSize = 32768;
 // Header is checksum (4 bytes), length (2 bytes), type (1 byte).
 static const int kHeaderSize = 4 + 2 + 1;
 
+
+/* 需要再研究下 */
+//4M 每次读取 vlog 的大小，GC 和 恢复 会用到
+//static const int vBlockSize = 4*1024*1024;
+
+// VlogHeader 只包含 length (4 bytes).
+static const int vHeaderSize = 4;
+// write_batch Header
+static const int wHeaderSize = 8 + 4 ;
+
 }  // namespace log
 }  // namespace leveldb
 
diff --git a/db/vlog_reader.cpp b/db/vlog_reader.cpp
new file mode 100644
index 0000000..4a928a3
--- /dev/null
+++ b/db/vlog_reader.cpp
@@ -0,0 +1,96 @@
+#include "vlog_reader.h"
+#include "leveldb/env.h"
+#include "util/coding.h"
+
+namespace leveldb {
+
+namespace log {
+
+VlogReader::VlogReader(SequentialFile *file, Reporter* reporter)
+    : file_(file),
+      file_random_(nullptr),
+      reporter_(reporter),
+      backing_store_(new char[kBlockSize]),
+      buffer_(),
+      eof_(false),
+      last_record_offset_(0) {}
+
+VlogReader::VlogReader(RandomAccessFile *file, Reporter* reporter)
+    : file_(nullptr),
+      file_random_(file),
+      reporter_(reporter),
+      backing_store_(new char[kBlockSize]),
+      buffer_(),
+      eof_(false),
+      last_record_offset_(0) {}
+
+VlogReader::~VlogReader() { delete[] backing_store_; }
+
+bool VlogReader::ReadValue(uint64_t offset, size_t length, Slice *key_value, char *scratch) {
+    if (file_random_ == nullptr) {
+        return false;
+    }
+    Status status = file_random_->Read(offset, length, key_value, scratch);
+    if (!status.ok()) {
+        return false;
+    }
+    return true;
+}
+
+bool VlogReader::ReadRecord(Slice *record, std::string *scratch) {
+    if (ReadPhysicalRecord(scratch)) {
+        *record = *scratch;
+        return true;
+    }
+    return false;
+}
+
+uint64_t VlogReader::LastRecordOffset() const {
+    return last_record_offset_;
+}
+
+void VlogReader::ReportCorruption(uint64_t bytes, const Status &reason) {
+    if (reporter_ != nullptr) {
+        reporter_->Corruption(static_cast<size_t>(bytes), reason);
+    }
+}
+
+bool VlogReader::ReadPhysicalRecord(std::string *result) {
+    result->clear();
+    buffer_.clear();
+
+    char* tmp_head = new char[vHeaderSize];
+    Status status = file_->Read(vHeaderSize, &buffer_, tmp_head);
+    if (!status.ok()) {
+        buffer_.clear();
+        ReportCorruption(kBlockSize, status);
+        eof_ = true;
+        return false;
+    } else if (buffer_.size() < vHeaderSize) {
+        eof_ = true;
+    }
+
+    if (!eof_) {
+        result->assign(buffer_.data(),buffer_.size());
+        uint32_t length = DecodeFixed32(buffer_.data());
+        buffer_.clear();
+        char* tmp = new char[length];
+        status = file_->Read(length, &buffer_, tmp);
+        if (status.ok() && buffer_.size() == length) {
+            *result += buffer_.ToString();
+        } else {
+            eof_ = true;
+        }
+        delete [] tmp;
+    }
+    delete [] tmp_head;
+
+    if (eof_) {
+        result->clear();
+        return false;
+    }
+    return true;
+}
+
+}  // namespace log
+}  // namespace leveldb
diff --git a/db/vlog_reader.h b/db/vlog_reader.h
new file mode 100644
index 0000000..b1d9484
--- /dev/null
+++ b/db/vlog_reader.h
@@ -0,0 +1,58 @@
+#ifndef LEVELDB_VLOG_READER_H
+#define LEVELDB_VLOG_READER_H
+
+#include <cstdint>
+
+#include "db/log_format.h"
+#include "leveldb/slice.h"
+#include "leveldb/status.h"
+
+namespace leveldb {
+
+class SequentialFile;
+class RandomAccessFile;
+
+namespace log {
+
+class VlogReader {
+    public:
+        class Reporter {
+            public:
+                virtual ~Reporter() = default;
+                virtual void Corruption(size_t bytes, const Status& status) = 0;
+        };
+
+        explicit VlogReader(SequentialFile* file, Reporter* reporter);
+        explicit VlogReader(RandomAccessFile* file, Reporter* reporter);
+
+        VlogReader(const VlogReader&) = delete;
+        VlogReader& operator=(const VlogReader&) = delete;
+
+        ~VlogReader();
+        // 从 vlog 中读取一条具体的值
+        bool ReadValue(uint64_t offset, size_t length, Slice* key_value, char* scratch);
+        bool ReadRecord(Slice* record, std::string* scratch);
+        // 返回 ReadRecord 函数读取的最后一条 record 的偏移（last_record_offset_）
+        uint64_t LastRecordOffset() const;
+
+    private:
+        /* 读取 vlog 物理记录（ data 部分） */
+        bool ReadPhysicalRecord(std::string* result);
+
+        void ReportCorruption(uint64_t bytes, const Status &reason);
+
+        SequentialFile* const file_;
+        RandomAccessFile* const file_random_;
+        Reporter* const reporter_;
+        char* const backing_store_;
+        Slice buffer_;
+        bool eof_;  // Last Read() indicated EOF by returning < kBlockSize
+
+        // Offset of the last record returned by ReadRecord.
+        uint64_t last_record_offset_;
+};
+
+}  // namespace log
+}  // namespace leveldb
+
+#endif  // LEVELDB_VLOG_READER_H
diff --git a/include/leveldb/options.h b/include/leveldb/options.h
index d755f46..e110c80 100644
--- a/include/leveldb/options.h
+++ b/include/leveldb/options.h
@@ -145,6 +145,16 @@ struct LEVELDB_EXPORT Options {
   // Many applications will benefit from passing the result of
   // NewBloomFilterPolicy() here.
   const FilterPolicy* filter_policy = nullptr;
+
+  /* 需要再研究下 */
+  // value log 的文件大小
+  uint64_t max_value_log_size = 500 * 1024 * 1024;
+  // gc 的回收阈值。
+  uint64_t garbage_collection_threshold = max_value_log_size / 4;
+  // gc 后台回收时候重新put的时候，默认的kv分离的值。
+  uint64_t background_garbage_collection_separate_ = 1024 * 1024 - 1;
+  // 在open 数据库的时候就进行全盘的log文件回收
+  bool start_garbage_collection = true;
 };
 
 // Options that control read operations
diff --git a/include/leveldb/status.h b/include/leveldb/status.h
index e327314..e4e6897 100644
--- a/include/leveldb/status.h
+++ b/include/leveldb/status.h
@@ -71,6 +71,12 @@ class LEVELDB_EXPORT Status {
   // Returns true iff the status indicates an InvalidArgument.
   bool IsInvalidArgument() const { return code() == kInvalidArgument; }
 
+  bool IsSeparated() const { return code() == kSeparated; }
+
+  void SetSeparated()  { code_ = kSeparated; }
+
+  void SetNotSeparated()  { code_ = kNotSeparated; }
+
   // Return a string representation of this status suitable for printing.
   // Returns the string "OK" for success.
   std::string ToString() const;
@@ -82,9 +88,12 @@ class LEVELDB_EXPORT Status {
     kCorruption = 2,
     kNotSupported = 3,
     kInvalidArgument = 4,
-    kIOError = 5
+    kIOError = 5,
+    kSeparated = 6,
+    kNotSeparated = 7
   };
 
+  Code code_;
   Code code() const {
     return (state_ == nullptr) ? kOk : static_cast<Code>(state_[4]);
   }
@@ -101,6 +110,7 @@ class LEVELDB_EXPORT Status {
 };
 
 inline Status::Status(const Status& rhs) {
+  code_ = rhs.code_;
   state_ = (rhs.state_ == nullptr) ? nullptr : CopyState(rhs.state_);
 }
 inline Status& Status::operator=(const Status& rhs) {
@@ -108,12 +118,14 @@ inline Status& Status::operator=(const Status& rhs) {
   // and the common case where both rhs and *this are ok.
   if (state_ != rhs.state_) {
     delete[] state_;
+    code_ = rhs.code_;
     state_ = (rhs.state_ == nullptr) ? nullptr : CopyState(rhs.state_);
   }
   return *this;
 }
 inline Status& Status::operator=(Status&& rhs) noexcept {
   std::swap(state_, rhs.state_);
+  std::swap(code_ , rhs.code_);
   return *this;
 }