Merge branch '3.0' of github.com:taosdata/tdengine into 3.0

CMakeLists.txt

@@ -50,6 +50,11 @@ if(${BUILD_WITH_LUCENE})
     cat("${CMAKE_SUPPORT_DIR}/lucene_CMakeLists.txt.in" ${DEPS_TMP_FILE})
 endif(${BUILD_WITH_LUCENE})
 
+## NuRaft
+if(${BUILD_WITH_NURAFT})
+    cat("${CMAKE_SUPPORT_DIR}/nuraft_CMakeLists.txt.in" ${DEPS_TMP_FILE})
+endif(${BUILD_WITH_NURAFT})
+
 ## download dependencies
 configure_file(${DEPS_TMP_FILE} "${CMAKE_SOURCE_DIR}/deps/deps-download/CMakeLists.txt")
 execute_process(COMMAND "${CMAKE_COMMAND}" -G "${CMAKE_GENERATOR}" .
@@ -69,4 +74,7 @@ target_include_directories(api INTERFACE "include/client")
 # src
 add_subdirectory(source)
 
+# docs 
+add_subdirectory(docs)
+
 # tests (TODO)

Jenkinsfile

@@ -142,7 +142,7 @@ pipeline {
           }
       }
   //     stage('Parallel test stage') {
-  //       //only build pr
+  //     skip defaultCheckout 
   //       options { skipDefaultCheckout() } 
   //       when {
   //         allOf{

cmake/cmake.options

@@ -16,7 +16,7 @@ option(
 option(
     BUILD_WITH_ROCKSDB 
     "If build with rocksdb" 
-    OFF
+    ON
 )
 
 option(
@@ -25,8 +25,20 @@ option(
     OFF
 )
 
+option(
+    BUILD_WITH_NURAFT
+    "If build with NuRaft" 
+    OFF
+)
+
 option(
     BUILD_DEPENDENCY_TESTS 
     "If build dependency tests" 
     OFF
+)
+
+option(
+    BUILD_DOCS
+    "If use doxygen build documents" 
+    ON
 )

cmake/nuraft_CMakeLists.txt.in

@@ -0,0 +1,12 @@
+
+# NuRaft
+ExternalProject_Add(NuRaft
+        GIT_REPOSITORY https://github.com/eBay/NuRaft.git
+        GIT_TAG v1.3.0
+        SOURCE_DIR "${CMAKE_SOURCE_DIR}/deps/nuraft"
+        BINARY_DIR "${CMAKE_SOURCE_DIR}/deps/nuraft"
+        CONFIGURE_COMMAND "./prepare.sh"
+        BUILD_COMMAND ""
+        INSTALL_COMMAND ""
+        TEST_COMMAND ""
+        )

deps/CMakeLists.txt

@@ -67,6 +67,12 @@ if(${BUILD_WITH_LUCENE})
     add_subdirectory(lucene)
 endif(${BUILD_WITH_LUCENE})
 
+# NuRaft
+if(${BUILD_WITH_NURAFT})
+    add_subdirectory(nuraft)
+endif(${BUILD_WITH_NURAFT})
+
+
 # ================================================================================================
 # DEPENDENCY TEST
 # ================================================================================================

docs/CMakeLists.txt

@@ -0,0 +1,23 @@
+# Generate API documentation
+## https://vicrucann.github.io/tutorials/quick-cmake-doxygen/
+if(${BUILD_DOCS})
+    find_package(Doxygen)
+    if (DOXYGEN_FOUND)
+        # Build the doc
+        set(DOXYGEN_IN ${CMAKE_SOURCE_DIR}/docs/Doxyfile.in)
+        set(DOXYGEN_OUT ${CMAKE_BINARY_DIR}/Doxyfile)
+
+        configure_file(${DOXYGEN_IN} ${DOXYGEN_OUT} @ONLY)
+        message("Doxygen build start")
+
+        add_custom_target(
+            tdengine_doxygen ALL
+            COMMAND ${DOXYGEN_EXECUTABLE} ${DOXYGEN_OUT}
+            WORKING_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR}
+            COMMENT "Generating API doxumentation with Doxygen"
+            VERBATIM
+        )
+    else(DOXYGEN_FOUND)
+        message("Doxygen need to be installed to generate the doxygen documentation")
+    endif(DOXYGEN_FOUND)
+endif(${BUILD_DOCS})

include/common/trequest.h

@@ -0,0 +1,47 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef _TD_TREQUEST_H_
+#define _TD_TREQUEST_H_
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/* ------------------------ TYPES EXPOSED ------------------------ */
+typedef struct SRequest      SRequest;
+typedef struct SReqBatch     SReqBatch;
+typedef struct SReqBatchIter SReqBatchIter;
+
+// SRequest
+
+// SReqBatch
+
+// SReqBatchIter
+void            tdInitRBIter(SReqBatchIter *pIter, SReqBatch *pReqBatch);
+const SRequest *tdRBIterNext(SReqBatchIter *pIter);
+void            tdClearRBIter(SReqBatchIter *pIter);
+
+/* ------------------------ TYPES DEFINITION ------------------------ */
+struct SReqBatchIter {
+  int        iReq;
+  SReqBatch *pReqBatch;
+};
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /*_TD_TREQUEST_H_*/

include/libs/sync/sync.h

@@ -26,13 +26,13 @@ extern "C" {
 typedef int32_t  SyncNodeId;
 typedef int32_t  SyncGroupId;
 typedef int64_t  SyncIndex;
-typedef uint64_t SSyncTerm;
+typedef uint64_t SyncTerm;
 
 typedef enum {
-  TAOS_SYNC_ROLE_FOLLOWER = 0,
-  TAOS_SYNC_ROLE_CANDIDATE = 1,
-  TAOS_SYNC_ROLE_LEADER = 2,
-} ESyncRole;
+  TAOS_SYNC_STATE_FOLLOWER = 0,
+  TAOS_SYNC_STATE_CANDIDATE = 1,
+  TAOS_SYNC_STATE_LEADER = 2,
+} ESyncState;
 
 typedef struct {
   void*  data;
@@ -55,19 +55,19 @@ typedef struct {
   int32_t   selfIndex;
   int32_t   replica;
   SNodeInfo node[TSDB_MAX_REPLICA];
-  ESyncRole role[TSDB_MAX_REPLICA];
+  ESyncState role[TSDB_MAX_REPLICA];
 } SNodesRole;
 
 typedef struct SSyncFSM {
   void* pData;
 
-  // apply committed log, bufs will be free by raft module
+  // apply committed log, bufs will be free by sync module
   int32_t (*applyLog)(struct SSyncFSM* fsm, SyncIndex index, const SSyncBuffer* buf, void* pData);
 
   // cluster commit callback
   int32_t (*onClusterChanged)(struct SSyncFSM* fsm, const SSyncCluster* cluster, void* pData);
 
-  // fsm return snapshot in ppBuf, bufs will be free by raft module
+  // fsm return snapshot in ppBuf, bufs will be free by sync module
   // TODO: getSnapshot SHOULD be async?
   int32_t (*getSnapshot)(struct SSyncFSM* fsm, SSyncBuffer** ppBuf, int32_t* objId, bool* isLast);
 
@@ -89,19 +89,30 @@ typedef struct SSyncLogStore {
   // write log with given index
   int32_t (*logWrite)(struct SSyncLogStore* logStore, SyncIndex index, SSyncBuffer* pBuf);
 
+  /** 
+   * read log from given index(included) with limit, return the actual num in nBuf,
+   * pBuf will be free in sync module
+   **/
+  int32_t (*logRead)(struct SSyncLogStore* logStore, SyncIndex index, int limit,
+                      SSyncBuffer* pBuf, int* nBuf);
+
   // mark log with given index has been commtted
   int32_t (*logCommit)(struct SSyncLogStore* logStore, SyncIndex index);
 
-  // prune log before given index
+  // prune log before given index(not included)
   int32_t (*logPrune)(struct SSyncLogStore* logStore, SyncIndex index);
 
-  // rollback log after given index
+  // rollback log after given index(included)
   int32_t (*logRollback)(struct SSyncLogStore* logStore, SyncIndex index);
+
+  // return last index of log
+  SyncIndex (*logLastIndex)(struct SSyncLogStore* logStore);
 } SSyncLogStore;
 
 typedef struct SSyncServerState {
   SyncNodeId voteFor;
-  SSyncTerm  term;
+  SyncTerm  term;
+  SyncIndex  commitIndex;
 } SSyncServerState;
 
 typedef struct SSyncClusterConfig {
@@ -122,9 +133,9 @@ typedef struct SStateManager {
 
   int32_t (*readServerState)(struct SStateManager* stateMng, SSyncServerState* state);
 
-  // void (*saveCluster)(struct SStateManager* stateMng, const SSyncClusterConfig* cluster);
+  void (*saveCluster)(struct SStateManager* stateMng, const SSyncClusterConfig* cluster);
 
-  // const SSyncClusterConfig* (*readCluster)(struct SStateManager* stateMng);
+  const SSyncClusterConfig* (*readCluster)(struct SStateManager* stateMng);
 } SStateManager;
 
 typedef struct {
@@ -146,13 +157,13 @@ SSyncNode* syncStart(const SSyncInfo*);
 void       syncReconfig(const SSyncNode*, const SSyncCluster*);
 void       syncStop(const SSyncNode*);
 
-int32_t syncPropose(SSyncNode* syncNode, SSyncBuffer buffer, void* pData, bool isWeak);
+int32_t syncPropose(SSyncNode* syncNode, const SSyncBuffer* pBuf, void* pData, bool isWeak);
 
-// int32_t syncAddNode(SSyncNode syncNode, const SNodeInfo *pNode);
+int32_t syncAddNode(SSyncNode syncNode, const SNodeInfo *pNode);
 
-// int32_t syncRemoveNode(SSyncNode syncNode, const SNodeInfo *pNode);
+int32_t syncRemoveNode(SSyncNode syncNode, const SNodeInfo *pNode);
 
-extern int32_t syncDebugFlag;
+extern int32_t sDebugFlag;
 
 #ifdef __cplusplus
 }

include/os/osDir.h

@@ -22,7 +22,7 @@ extern "C" {
 
 void taosRemoveDir(const char *dirname);
 bool taosDirExist(char *dirname);
-bool taosMkDir(char *dirname);
+bool taosMkDir(const char *dirname);
 void taosRemoveOldFiles(char *dirname, int32_t keepDays);
 bool taosExpandDir(char *dirname, char *outname, int32_t maxlen);
 bool taosRealPath(char *dirname, int32_t maxlen);

include/server/vnode/meta/impl/metaImpl.h

@@ -0,0 +1,75 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef _TD_META_IMPL_H_
+#define _TD_META_IMPL_H_
+
+#include "os.h"
+
+#include "taosmsg.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+typedef uint64_t tb_uid_t;
+
+/* ------------------------ SMetaOptions ------------------------ */
+struct SMetaOptions {
+  size_t lruCacheSize;  // LRU cache size
+};
+
+/* ------------------------ STbOptions ------------------------ */
+#define META_NORMAL_TABLE ((uint8_t)1)
+#define META_SUPER_TABLE ((uint8_t)2)
+#define META_CHILD_TABLE ((uint8_t)3)
+
+typedef struct {
+} SSMAOptions;
+
+// super table options
+typedef struct {
+  tb_uid_t  uid;
+  STSchema* pSchema;
+  STSchema* pTagSchema;
+} SSTbOptions;
+
+// child table options
+typedef struct {
+  tb_uid_t suid;
+  SKVRow   tags;
+} SCTbOptions;
+
+// normal table options
+typedef struct {
+  STSchema* pSchame;
+} SNTbOptions;
+
+struct STbOptions {
+  uint8_t     type;
+  char*       name;
+  uint32_t    ttl;   // time to live in (SECONDS)
+  SSMAOptions bsma;  // Block-wise sma
+  union {
+    SSTbOptions stbOptions;
+    SNTbOptions ntbOptions;
+    SCTbOptions ctbOptions;
+  };
+};
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /*_TD_META_IMPL_H_*/

include/server/vnode/meta/meta.h

@@ -16,81 +16,36 @@
 #ifndef _TD_META_H_
 #define _TD_META_H_
 
-#include "taosmsg.h"
-
-#include "os.h"
+#include "impl/metaImpl.h"
 
 #ifdef __cplusplus
 extern "C" {
 #endif
 
-/* ------------------------ APIs Exposed ------------------------ */
-
 // Types exported
-typedef uint64_t                tb_uid_t;
-typedef struct SMeta            SMeta;
-typedef struct SMetaOpts        SMetaOpts;
-typedef struct SMetaQueryHandle SMetaQueryHandle;
-typedef struct SMetaQueryOpts   SMetaQueryOpts;
-typedef struct STableOpts       STableOpts;
+typedef struct SMeta        SMeta;
+typedef struct SMetaOptions SMetaOptions;
+typedef struct STbOptions   STbOptions;
 
 // SMeta operations
-int    metaCreate(const char *path);
-void   metaDestroy(const char *path);
-SMeta *metaOpen(SMetaOpts *);
+SMeta *metaOpen(const char *path, const SMetaOptions *);
 void   metaClose(SMeta *);
-int    metaCreateTable(SMeta *, const STableOpts *);
-int    metaDropTable(SMeta *, uint64_t tuid_t);
-int    metaAlterTable(SMeta *, void *);
+void   metaRemove(const char *path);
+int    metaCreateTable(SMeta *pMeta, const STbOptions *);
+int    metaDropTable(SMeta *pMeta, tb_uid_t uid);
 int    metaCommit(SMeta *);
 
 // Options
-SMetaOpts *metaOptionsCreate();
-void       metaOptionsDestroy(SMetaOpts *);
-void       metaOptionsSetCache(SMetaOpts *, size_t capacity);
-
-// SMetaQueryHandle
-SMetaQueryHandle *metaQueryHandleCreate(SMetaQueryOpts *);
-void              metaQueryHandleDestroy(SMetaQueryHandle *);
-
-// SMetaQueryOpts
-SMetaQueryOpts *metaQueryOptionsCreate();
-void            metaQueryOptionsDestroy(SMetaQueryOpts *);
+void metaOptionsInit(SMetaOptions *);
+void metaOptionsClear(SMetaOptions *);
 
 // STableOpts
 #define META_TABLE_OPTS_DECLARE(name) STableOpts name = {0}
-void metaNormalTableOptsInit(STableOpts *, const char *name, const STSchema *pSchema);
-void metaSuperTableOptsInit(STableOpts *, const char *name, tb_uid_t uid, const STSchema *pSchema,
+void metaNormalTableOptsInit(STbOptions *, const char *name, const STSchema *pSchema);
+void metaSuperTableOptsInit(STbOptions *, const char *name, tb_uid_t uid, const STSchema *pSchema,
                             const STSchema *pTagSchema);
-void metaChildTableOptsInit(STableOpts *, const char *name, tb_uid_t suid, const SKVRow tags);
-void metaTableOptsClear(STableOpts *);
-
-/* ------------------------ Impl should hidden ------------------------ */
-typedef enum { META_INIT_TABLE = 0, META_SUPER_TABLE = 1, META_CHILD_TABLE = 2, META_NORMAL_TABLE = 3 } EMetaTableT;
-typedef struct SSuperTableOpts {
-  tb_uid_t  uid;
-  STSchema *pSchema;     // (ts timestamp, a int)
-  STSchema *pTagSchema;  // (tag1 binary(10), tag2 int)
-} SSuperTableOpts;
-
-typedef struct SChildTableOpts {
-  tb_uid_t suid;  // super table uid
-  SKVRow   tags;  // tag value of the child table
-} SChildTableOpts;
-
-typedef struct SNormalTableOpts {
-  STSchema *pSchema;
-} SNormalTableOpts;
-
-struct STableOpts {
-  int8_t type;
-  char * name;
-  union {
-    SSuperTableOpts  superOpts;
-    SChildTableOpts  childOpts;
-    SNormalTableOpts normalOpts;
-  };
-};
+void metaChildTableOptsInit(STbOptions *, const char *name, tb_uid_t suid, const SKVRow tags);
+void metaTableOptsClear(STbOptions *);
 
 #ifdef __cplusplus
 }

include/server/vnode/tsdb/impl/tsdbImpl.h

@@ -23,6 +23,7 @@ extern "C" {
 #endif
 
 struct STsdbOptions {
+  size_t lruCacheSize;
   /* TODO */
 };
 

include/server/vnode/vnode.h

@@ -17,14 +17,80 @@
 #define _TD_VNODE_H_
 
 #include "os.h"
-#include "taosmsg.h"
-#include "trpc.h"
+#include "trequest.h"
+
+#include "meta.h"
+#include "tq.h"
+#include "tsdb.h"
 
 #ifdef __cplusplus
 extern "C" {
 #endif
 
-typedef struct SVnode SVnode;
+/* ------------------------ TYPES EXPOSED ------------------------ */
+typedef struct SVnode        SVnode;
+typedef struct SVnodeOptions SVnodeOptions;
+
+/* ------------------------ SVnode ------------------------ */
+SVnode *vnodeOpen(const char *path, const SVnodeOptions *pVnodeOptions);
+void    vnodeClose(SVnode *pVnode);
+void    vnodeDestroy(const char *path);
+int     vnodeProcessWriteReqs(SVnode *pVnode, SReqBatch *pReqBatch);
+int     vnodeApplyWriteRequest(SVnode *pVnode, const SRequest *pRequest);
+int     vnodeProcessReadReq(SVnode *pVnode, SRequest *pReq);
+int     vnodeProcessSyncReq(SVnode *pVnode, SRequest *pReq);
+
+/* ------------------------ SVnodeOptions ------------------------ */
+void vnodeOptionsInit(SVnodeOptions *);
+void vnodeOptionsClear(SVnodeOptions *);
+
+/* ------------------------ STRUCT DEFINITIONS ------------------------ */
+struct SVnodeOptions {
+  /**
+   * @brief write buffer size in BYTES
+   *
+   */
+  uint64_t wsize;
+
+  /**
+   * @brief time to live of tables in this vnode
+   * in SECONDS
+   *
+   */
+  uint32_t ttl;
+
+  /**
+   * @brief if time-series requests eventual consistency
+   *
+   */
+  bool isWeak;
+
+  /**
+   * @brief if the allocator is heap allcator or arena allocator
+   *
+   */
+  bool isHeapAllocator;
+
+  /**
+   * @brief TSDB options
+   *
+   */
+  STsdbOptions tsdbOptions;
+
+  /**
+   * @brief META options
+   *
+   */
+  SMetaOptions metaOptions;
+  // STqOptions   tqOptions; // TODO
+};
+
+/* ------------------------ FOR COMPILE ------------------------ */
+
+#if 1
+
+#include "taosmsg.h"
+#include "trpc.h"
 
 typedef struct {
   char     db[TSDB_FULL_DB_NAME_LEN];
@@ -71,8 +137,6 @@ typedef struct {
 int32_t vnodeInit(SVnodePara);
 void    vnodeCleanup();
 
-SVnode *vnodeOpen(int32_t vgId, const char *path);
-void    vnodeClose(SVnode *pVnode);
 int32_t vnodeAlter(SVnode *pVnode, const SVnodeCfg *pCfg);
 SVnode *vnodeCreate(int32_t vgId, const char *path, const SVnodeCfg *pCfg);
 void    vnodeDrop(SVnode *pVnode);
@@ -86,6 +150,8 @@ int32_t    vnodeAppendMsg(SVnodeMsg *pMsg, SRpcMsg *pRpcMsg);
 void       vnodeCleanupMsg(SVnodeMsg *pMsg);
 void       vnodeProcessMsg(SVnode *pVnode, SVnodeMsg *pMsg, EVnMsgType msgType);
 
+#endif
+
 #ifdef __cplusplus
 }
 #endif

include/util/amalloc.h

@@ -1,52 +0,0 @@
-/*
- * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
- *
- * This program is free software: you can use, redistribute, and/or modify
- * it under the terms of the GNU Affero General Public License, version 3
- * or later ("AGPL"), as published by the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.
- *
- * You should have received a copy of the GNU Affero General Public License
- * along with this program. If not, see <http://www.gnu.org/licenses/>.
- */
-
-#ifndef _TD_AMALLOC_H_
-#define _TD_AMALLOC_H_
-
-#include "os.h"
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-#define AMALLOC_APIS                                  \
-  void *(*malloc)(void *, size_t size);               \
-  void *(*calloc)(void *, size_t nmemb, size_t size); \
-  void *(*realloc)(void *, size_t size);              \
-  void (*free)(void *ptr);
-
-// Interfaces to implement
-typedef struct {
-  AMALLOC_APIS
-} SMemAllocatorIf;
-
-typedef struct {
-  void *impl;
-  AMALLOC_APIS
-} SMemAllocator;
-
-#define amalloc(allocator, size) ((allocator) ? (*((allocator)->malloc))((allocator)->impl, (size)) : malloc(size))
-#define acalloc(allocator, nmemb, size) \
-  ((allocator) ? (*((allocator)->calloc))((allocator)->impl, (nmemb), (size)) : calloc((nmemb), (size)))
-#define arealloc(allocator, ptr, size) \
-  ((allocator) ? (*((allocator)->realloc))((allocator)->impl, (ptr), (size)) : realloc((ptr), (size)))
-#define afree(allocator, ptr, size) ((allocator) ? (*((allocator)->free))((allocator)->impl, (ptr), (size)) : free(ptr))
-
-#ifdef __cplusplus
-}
-#endif
-
-#endif /*_TD_AMALLOC_H_*/

include/util/mallocator.h

@@ -0,0 +1,63 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef _TD_MALLOCATOR_H_
+#define _TD_MALLOCATOR_H_
+
+#include "os.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+typedef struct SMemAllocator SMemAllocator;
+
+#define MALLOCATOR_APIS                                        \
+  void *(*malloc)(SMemAllocator *, size_t size);               \
+  void *(*calloc)(SMemAllocator *, size_t nmemb, size_t size); \
+  void *(*realloc)(SMemAllocator *, void *ptr, size_t size);   \
+  void (*free)(SMemAllocator *, void *ptr);                    \
+  size_t (*usage)(SMemAllocator *);
+
+// Interfaces to implement
+typedef struct {
+  MALLOCATOR_APIS
+} SMemAllocatorIf;
+
+struct SMemAllocator {
+  void * impl;
+  size_t usize;
+  MALLOCATOR_APIS
+};
+
+// heap allocator
+SMemAllocator *tdCreateHeapAllocator();
+void           tdDestroyHeapAllocator(SMemAllocator *pMemAllocator);
+
+// arena allocator
+SMemAllocator *tdCreateArenaAllocator(size_t size);
+void           tdDestroyArenaAllocator(SMemAllocator *);
+
+#define mMalloc(pMemAllocator, size) (*(pMemAllocator->malloc))(pMemAllocator, size)
+#define mCalloc(pMemAllocator, nmemb, size) (*(pMemAllocator->calloc))(pMemAllocator, nmemb, size)
+#define mRealloc(pMemAllocator, ptr, size) (*(pMemAllocator->realloc))(pMemAllocator, ptr, size)
+#define mFree(pMemAllocator, ptr) (*(pMemAllocator->free))(pMemAllocator, ptr)
+#define mUsage(pMemAllocator) (*(pMemAllocator->usage))(pMemAllocator)
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /*_TD_MALLOCATOR_H_*/

source/dnode/mgmt/src/dnodeVnodes.c

@@ -376,7 +376,7 @@ static void *dnodeOpenVnodeFunc(void *param) {
 
     char path[PATH_MAX + 20] = {0};
     snprintf(path, sizeof(path),"%s/vnode%d", tsVnodeDir, pVnode->vgId);
-    SVnode *pImpl = vnodeOpen(pVnode->vgId, path);
+    SVnode *pImpl = vnodeOpen(path, NULL);
     if (pImpl == NULL) {
       dError("vgId:%d, failed to open vnode by thread:%d", pVnode->vgId, pThread->threadIndex);
       pThread->failed++;

source/dnode/vnode/impl/CMakeLists.txt

@@ -7,6 +7,7 @@ target_include_directories(
 )
 target_link_libraries(
     vnode
+    PUBLIC os
     PUBLIC transport
     PUBLIC meta
     PUBLIC tq

source/dnode/vnode/impl/inc/vnodeAllocatorPool.h

@@ -0,0 +1,39 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef _TD_VNODE_ALLOCATOR_POOL_H_
+#define _TD_VNODE_ALLOCATOR_POOL_H_
+
+#include "vnode.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+typedef struct {
+  int            nexta;
+  int            enda;
+  SMemAllocator *free[3];
+  SMemAllocator *used[3];
+} SVAllocatorPool;
+
+int  vnodeOpenAllocatorPool(SVnode *pVnode);
+void vnodeCloseAllocatorPool(SVnode *pVnode);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /*_TD_VNODE_ALLOCATOR_POOL_H_*/

source/dnode/vnode/impl/inc/vnodeCommit.h

@@ -16,13 +16,14 @@
 #ifndef _TD_VNODE_COMMIT_H_
 #define _TD_VNODE_COMMIT_H_
 
-#include "vnodeInt.h"
+#include "vnode.h"
 
 #ifdef __cplusplus
 extern "C" {
 #endif
 
-int vnodeAsyncCommit(SVnode *pVnode);
+bool vnodeShouldCommit(SVnode *pVnode);
+int  vnodeAsyncCommit(SVnode *pVnode);
 
 #ifdef __cplusplus
 }

source/dnode/vnode/impl/inc/vnodeDef.h

@@ -0,0 +1,52 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef _TD_VNODE_DEF_H_
+#define _TD_VNODE_DEF_H_
+
+#include "mallocator.h"
+#include "sync.h"
+#include "tlockfree.h"
+
+#include "vnode.h"
+#include "vnodeAllocatorPool.h"
+#include "vnodeCommit.h"
+#include "vnodeFileSystem.h"
+#include "vnodeOptions.h"
+#include "vnodeStateMgr.h"
+#include "vnodeSync.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+struct SVnode {
+  char*            path;
+  SVnodeOptions    options;
+  SVState          state;
+  SVAllocatorPool* pool;
+  SMemAllocator*   inuse;
+  SMeta*           pMeta;
+  STsdb*           pTsdb;
+  STQ*             pTq;
+  SVnodeSync*      pSync;
+  SVnodeFS*        pFs;
+};
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /*_TD_VNODE_DEF_H_*/

source/dnode/vnode/impl/inc/vnodeFileSystem.h

@@ -0,0 +1,30 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef _TD_VNODE_FILE_SYSTEM_H_
+#define _TD_VNODE_FILE_SYSTEM_H_
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+typedef struct {
+} SVnodeFS;
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /*_TD_VNODE_FILE_SYSTEM_H_*/

source/dnode/vnode/impl/inc/vnodeInt.h

@@ -18,7 +18,6 @@
 
 #include "vnode.h"
 
-#include "amalloc.h"
 #include "meta.h"
 #include "sync.h"
 #include "tlog.h"
@@ -39,16 +38,6 @@ extern int32_t vDebugFlag;
 #define vDebug(...) { if (vDebugFlag & DEBUG_DEBUG) { taosPrintLog("VND ", vDebugFlag, __VA_ARGS__); }}
 #define vTrace(...) { if (vDebugFlag & DEBUG_TRACE) { taosPrintLog("VND ", vDebugFlag, __VA_ARGS__); }}
 
-typedef struct SVnode {
-  int32_t    vgId;
-  SVnodeCfg  cfg;
-  SMeta     *pMeta;
-  STsdb     *pTsdb;
-  STQ       *pTQ;
-  SWal      *pWal;
-  SSyncNode *pSync;
-} SVnode;
-
 #ifdef __cplusplus
 }
 #endif

source/dnode/vnode/impl/inc/vnodeMemAllocator.h

@@ -16,40 +16,12 @@
 #ifndef _TD_VNODE_MEM_ALLOCATOR_H_
 #define _TD_VNODE_MEM_ALLOCATOR_H_
 
-#include "vnodeInt.h"
-
 #ifdef __cplusplus
 extern "C" {
 #endif
 
-typedef struct SVnodeMemAllocator SVnodeMemAllocator;
-
-SVnodeMemAllocator *VMACreate(size_t size /* base size */, size_t ssize /* step size */,
-                              size_t threshold /* threshold size when full*/);
-void                VMADestroy(SVnodeMemAllocator *pvma);
-void                VMAReset(SVnodeMemAllocator *pvma);
-void *              VMAMalloc(SVnodeMemAllocator *pvma, size_t size);
-void                VMAFree(SVnodeMemAllocator *pvma, void *ptr);
-bool                VMAIsFull(SVnodeMemAllocator *pvma);
-
-// ------------------ FOR TEST ONLY ------------------
-typedef struct SVMANode {
-  struct SVMANode *prev;
-  size_t           tsize;
-  size_t           used;
-  char             data[];
-} SVMANode;
-
-struct SVnodeMemAllocator {
-  bool      full;       // if allocator is full
-  size_t    threshold;  // threshold;
-  size_t    ssize;      // step size to allocate
-  SVMANode *inuse;      // inuse node to allocate
-  SVMANode  node;       // basic node to use
-};
-
 #ifdef __cplusplus
 }
 #endif
 
-#endif /*_TD_VNODE_MEM_ALLOCATOR_H_*/
+#endif /*_TD_VNODE_MEM_ALLOCATOR_H_*/

source/dnode/vnode/impl/inc/vnodeOptions.h

@@ -0,0 +1,34 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef _TD_VNODE_OPTIONS_H_
+#define _TD_VNODE_OPTIONS_H_
+
+#include "vnode.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+extern const SVnodeOptions defaultVnodeOptions;
+
+int  vnodeValidateOptions(const SVnodeOptions *);
+void vnodeOptionsCopy(SVnodeOptions *pDest, const SVnodeOptions *pSrc);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /*_TD_VNODE_OPTIONS_H_*/

source/dnode/vnode/impl/inc/vnodeRequest.h

@@ -0,0 +1,27 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef _TD_VNODE_REQUEST_H_
+#define _TD_VNODE_REQUEST_H_
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /*_TD_VNODE_REQUEST_H_*/

source/dnode/vnode/impl/inc/vnodeStateMgr.h

@@ -0,0 +1,30 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef _TD_VNODE_STATE_MGR_H_
+#define _TD_VNODE_STATE_MGR_H_
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+typedef struct {
+} SVState;
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /*_TD_VNODE_STATE_MGR_H_*/

source/dnode/vnode/impl/inc/vnodeSync.h

@@ -0,0 +1,33 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef _TD_VNODE_SYNC_H_
+#define _TD_VNODE_SYNC_H_
+
+#include "sync.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+typedef struct {
+  /* data */
+} SVnodeSync;
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /*_TD_VNODE_SYNC_H_*/

source/dnode/vnode/impl/inc/vnodeWrite.h

@@ -16,10 +16,11 @@
 #ifndef _TD_VNODE_WRITE_H_
 #define _TD_VNODE_WRITE_H_
 
+#include "vnode.h"
+
 #ifdef __cplusplus
 extern "C" {
 #endif
-#include "vnodeInt.h"
 
 void vnodeProcessWriteMsg(SVnode* pVnode, SVnodeMsg* pMsg);
 

source/dnode/vnode/impl/src/vnodeAllocatorPool.c

@@ -0,0 +1,37 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "vnodeDef.h"
+
+int vnodeOpenAllocatorPool(SVnode *pVnode) {
+  // TODO
+  return 0;
+}
+
+void vnodeCloseAllocatorPool(SVnode *pVnode) {
+  if (pVnode->pool) {
+  }
+}
+
+/* ------------------------ STATIC METHODS ------------------------ */
+static SVAllocatorPool *vapCreate() {
+  SVAllocatorPool *pPool = NULL;
+  /* TODO */
+  return pPool;
+}
+
+static void vapDestroy() {
+  // TODO
+}

source/dnode/vnode/impl/src/vnodeCommit.c

@@ -13,13 +13,15 @@
  * along with this program. If not, see <http://www.gnu.org/licenses/>.
  */
 
-#include "vnodeInt.h"
+#include "vnodeDef.h"
 
 static int vnodeStartCommit(SVnode *pVnode);
 static int vnodeEndCommit(SVnode *pVnode);
 
+bool vnodeShouldCommit(SVnode *pVnode) { return false; }
+
 int vnodeAsyncCommit(SVnode *pVnode) {
-  #if 0
+#if 0
   if (vnodeStartCommit(pVnode) < 0) {
     // TODO
   }

source/dnode/vnode/impl/src/vnodeFileSystem.c

@@ -0,0 +1,16 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "vnodeDef.h"

source/dnode/vnode/impl/src/vnodeInt.c

@@ -20,8 +20,6 @@
 int32_t vnodeInit(SVnodePara para) { return 0; }
 void    vnodeCleanup() {}
 
-SVnode *vnodeOpen(int32_t vgId, const char *path) { return NULL; }
-void    vnodeClose(SVnode *pVnode) {}
 int32_t vnodeAlter(SVnode *pVnode, const SVnodeCfg *pCfg) { return 0; }
 SVnode *vnodeCreate(int32_t vgId, const char *path, const SVnodeCfg *pCfg) { return NULL; }
 void    vnodeDrop(SVnode *pVnode) {}
@@ -69,4 +67,4 @@ void vnodeProcessMsg(SVnode *pVnode, SVnodeMsg *pMsg, EVnMsgType msgType) {
     case VN_MSG_TYPE_FETCH:
       break;
   }
-}
+}

source/dnode/vnode/impl/src/vnodeMain.c

@@ -0,0 +1,125 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "vnodeDef.h"
+
+static SVnode *vnodeNew(const char *path, const SVnodeOptions *pVnodeOptions);
+static void    vnodeFree(SVnode *pVnode);
+static int     vnodeOpenImpl(SVnode *pVnode);
+static void    vnodeCloseImpl(SVnode *pVnode);
+
+SVnode *vnodeOpen(const char *path, const SVnodeOptions *pVnodeOptions) {
+  SVnode *pVnode = NULL;
+
+  // Set default options
+  if (pVnodeOptions == NULL) {
+    pVnodeOptions = &defaultVnodeOptions;
+  }
+
+  // Validate options
+  if (vnodeValidateOptions(pVnodeOptions) < 0) {
+    // TODO
+    return NULL;
+  }
+
+  // Create the handle
+  pVnode = vnodeNew(path, pVnodeOptions);
+  if (pVnode == NULL) {
+    // TODO: handle error
+    return NULL;
+  }
+
+  taosMkDir(path);
+
+  // Open the vnode
+  if (vnodeOpenImpl(pVnode) < 0) {
+    // TODO: handle error
+    return NULL;
+  }
+
+  return pVnode;
+}
+
+void vnodeClose(SVnode *pVnode) {
+  if (pVnode) {
+    vnodeCloseImpl(pVnode);
+    vnodeFree(pVnode);
+  }
+}
+
+void vnodeDestroy(const char *path) { taosRemoveDir(path); }
+
+/* ------------------------ STATIC METHODS ------------------------ */
+static SVnode *vnodeNew(const char *path, const SVnodeOptions *pVnodeOptions) {
+  SVnode *pVnode = NULL;
+
+  pVnode = (SVnode *)calloc(1, sizeof(*pVnode));
+  if (pVnode == NULL) {
+    // TODO
+    return NULL;
+  }
+
+  pVnode->path = strdup(path);
+  vnodeOptionsCopy(&(pVnode->options), pVnodeOptions);
+
+  return pVnode;
+}
+
+static void vnodeFree(SVnode *pVnode) {
+  if (pVnode) {
+    tfree(pVnode->path);
+    free(pVnode);
+  }
+}
+
+static int vnodeOpenImpl(SVnode *pVnode) {
+  char dir[TSDB_FILENAME_LEN];
+
+  // Open allocator pool
+  if (vnodeOpenAllocatorPool(pVnode) < 0) {
+    // TODO: handle error
+    return -1;
+  }
+
+  // Open meta
+  sprintf(dir, "%s/meta", pVnode->path);
+  pVnode->pMeta = metaOpen(dir, &(pVnode->options.metaOptions));
+  if (pVnode->pMeta == NULL) {
+    // TODO: handle error
+    return -1;
+  }
+
+  // Open tsdb
+  sprintf(dir, "%s/tsdb", pVnode->path);
+  pVnode->pTsdb = tsdbOpen(dir, &(pVnode->options.tsdbOptions));
+  if (pVnode->pTsdb == NULL) {
+    // TODO: handle error
+    return -1;
+  }
+
+  // TODO: Open TQ
+
+  // TODO
+  return 0;
+}
+
+static void vnodeCloseImpl(SVnode *pVnode) {
+  if (pVnode) {
+    vnodeCloseAllocatorPool(pVnode);
+    // TODO: Close TQ
+    tsdbClose(pVnode->pTsdb);
+    metaClose(pVnode->pMeta);
+  }
+}

source/dnode/vnode/impl/src/vnodeMemAllocator.c

@@ -13,112 +13,88 @@
  * along with this program. If not, see <http://www.gnu.org/licenses/>.
  */
 
-#include "vnodeMemAllocator.h"
+#include "vnodeDef.h"
 
-#define VMA_IS_FULL(pvma) \
-  (((pvma)->inuse != &((pvma)->node)) || ((pvma)->inuse->tsize - (pvma)->inuse->used < (pvma)->threshold))
+#define VNODE_HEAP_ALLOCATOR 0
+#define VNODE_ARENA_ALLOCATOR 1
 
-static SVMANode *VMANodeNew(size_t size);
-static void      VMANodeFree(SVMANode *node);
+typedef struct {
+  uint64_t tsize;
+  uint64_t used;
+} SVHeapAllocator;
 
-SVnodeMemAllocator *VMACreate(size_t size, size_t ssize, size_t threshold) {
-  SVnodeMemAllocator *pvma = NULL;
+typedef struct SVArenaNode {
+  struct SVArenaNode *prev;
+  void *              nptr;
+  char                data[];
+} SVArenaNode;
 
-  if (size < threshold) {
+typedef struct {
+  SVArenaNode *inuse;
+  SVArenaNode  node;
+} SVArenaAllocator;
+
+typedef struct {
+  int8_t   type;
+  uint64_t tsize;
+  T_REF_DECLARE()
+  union {
+    SVHeapAllocator  vha;
+    SVArenaAllocator vaa;
+  };
+} SVMemAllocator;
+
+SMemAllocator *vnodeCreateMemAllocator(int8_t type, uint64_t tsize, uint64_t ssize /* step size only for arena */) {
+  SMemAllocator * pma;
+  uint64_t        msize;
+  SVMemAllocator *pva;
+
+  msize = sizeof(*pma) + sizeof(SVMemAllocator);
+  if (type == VNODE_ARENA_ALLOCATOR) {
+    msize += tsize;
+  }
+
+  pma = (SMemAllocator *)calloc(1, msize);
+  if (pma == NULL) {
     return NULL;
   }
 
-  pvma = (SVnodeMemAllocator *)malloc(sizeof(*pvma) + size);
-  if (pvma) {
-    pvma->full = false;
-    pvma->threshold = threshold;
-    pvma->ssize = ssize;
-    pvma->inuse = &(pvma->node);
+  pma->impl = POINTER_SHIFT(pma, sizeof(*pma));
+  pva = (SVMemAllocator *)(pma->impl);
+  pva->type = type;
+  pva->tsize = tsize;
 
-    pvma->inuse->prev = NULL;
-    pvma->inuse->tsize = size;
-    pvma->inuse->used = 0;
+  if (type == VNODE_HEAP_ALLOCATOR) {
+    pma->malloc = NULL;
+    pma->calloc = NULL;
+    pma->realloc = NULL;
+    pma->free = NULL;
+    pma->usage = NULL;
+  } else if (type == VNODE_ARENA_ALLOCATOR) {
+    pma->malloc = NULL;
+    pma->calloc = NULL;
+    pma->realloc = NULL;
+    pma->free = NULL;
+    pma->usage = NULL;
+  } else {
+    ASSERT(0);
   }
 
-  return pvma;
+  return pma;
 }
 
-void VMADestroy(SVnodeMemAllocator *pvma) {
-  if (pvma) {
-    VMAReset(pvma);
-    free(pvma);
-  }
+void vnodeDestroyMemAllocator(SMemAllocator *pma) {
+  // TODO
 }
 
-void VMAReset(SVnodeMemAllocator *pvma) {
-  while (pvma->inuse != &(pvma->node)) {
-    SVMANode *node = pvma->inuse;
-    pvma->inuse = node->prev;
-    VMANodeFree(node);
-  }
-
-  pvma->inuse->used = 0;
-  pvma->full = false;
+void vnodeRefMemAllocator(SMemAllocator *pma) {
+  // TODO
 }
 
-void *VMAMalloc(SVnodeMemAllocator *pvma, size_t size) {
-  void * ptr = NULL;
-  size_t tsize = size + sizeof(size_t);
-
-  if (pvma->inuse->tsize - pvma->inuse->used < tsize) {
-    SVMANode *pNode = VMANodeNew(MAX(pvma->ssize, tsize));
-    if (pNode == NULL) {
-      return NULL;
-    }
-
-    pNode->prev = pvma->inuse;
-    pvma->inuse = pNode;
-  }
-
-  ptr = pvma->inuse->data + pvma->inuse->used;
-  pvma->inuse->used += tsize;
-  *(size_t *)ptr = size;
-  ptr = POINTER_SHIFT(ptr, sizeof(size_t));
-
-  pvma->full = VMA_IS_FULL(pvma);
-
-  return ptr;
+void vnodeUnrefMemAllocator(SMemAllocator *pma) {
+  // TODO
 }
 
-void VMAFree(SVnodeMemAllocator *pvma, void *ptr) {
-  if (ptr) {
-    size_t size = *(size_t *)POINTER_SHIFT(ptr, -sizeof(size_t));
-    if (POINTER_SHIFT(ptr, size) == pvma->inuse->data + pvma->inuse->used) {
-      pvma->inuse->used -= (size + sizeof(size_t));
+/* ------------------------ Heap Allocator IMPL ------------------------ */
 
-      if ((pvma->inuse->used == 0) && (pvma->inuse != &(pvma->node))) {
-        SVMANode *node = pvma->inuse;
-        pvma->inuse = node->prev;
-        VMANodeFree(node);
-      }
-
-      pvma->full = VMA_IS_FULL(pvma);
-    }
-  }
-}
-
-bool VMAIsFull(SVnodeMemAllocator *pvma) { return pvma->full; }
-
-static SVMANode *VMANodeNew(size_t size) {
-  SVMANode *node = NULL;
-
-  node = (SVMANode *)malloc(sizeof(*node) + size);
-  if (node) {
-    node->prev = NULL;
-    node->tsize = size;
-    node->used = 0;
-  }
-
-  return node;
-}
-
-static void VMANodeFree(SVMANode *node) {
-  if (node) {
-    free(node);
-  }
-}
+/* ------------------------ Arena Allocator IMPL ------------------------ */

source/dnode/vnode/impl/src/vnodeOptions.c

@@ -0,0 +1,34 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "vnodeDef.h"
+
+const SVnodeOptions defaultVnodeOptions = {0}; /* TODO */
+
+void vnodeOptionsInit(SVnodeOptions *pVnodeOptions) { /* TODO */
+  vnodeOptionsCopy(pVnodeOptions, &defaultVnodeOptions);
+}
+
+void vnodeOptionsClear(SVnodeOptions *pVnodeOptions) { /* TODO */
+}
+
+int vnodeValidateOptions(const SVnodeOptions *pVnodeOptions) {
+  // TODO
+  return 0;
+}
+
+void vnodeOptionsCopy(SVnodeOptions *pDest, const SVnodeOptions *pSrc) {
+  memcpy((void *)pDest, (void *)pSrc, sizeof(SVnodeOptions));
+}

source/dnode/vnode/impl/src/vnodeRead.c

@@ -13,5 +13,4 @@
  * along with this program. If not, see <http://www.gnu.org/licenses/>.
  */
 
-#define _DEFAULT_SOURCE
-#include "vnodeRead.h"
+#include "vnodeDef.h"

source/dnode/vnode/impl/src/vnodeRequest.c

@@ -0,0 +1,14 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */

source/dnode/vnode/impl/src/vnodeStateMgr.c

@@ -0,0 +1,14 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */

source/dnode/vnode/impl/src/vnodeSync.c

@@ -0,0 +1,14 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */

source/dnode/vnode/impl/src/vnodeWrite.c

@@ -13,5 +13,50 @@
  * along with this program. If not, see <http://www.gnu.org/licenses/>.
  */
 
-#define _DEFAULT_SOURCE
-#include "vnodeWrite.h"
+#include "vnodeDef.h"
+
+int vnodeProcessWriteReqs(SVnode *pVnode, SReqBatch *pReqBatch) {
+  /* TODO */
+  return 0;
+}
+
+int vnodeApplyWriteRequest(SVnode *pVnode, const SRequest *pRequest) {
+  int      reqType;        /* TODO */
+  size_t   reqSize;        /* TODO */
+  uint64_t reqVersion = 0; /* TODO */
+  int      code = 0;
+
+  // Copy the request to vnode buffer
+  SRequest *pReq = mMalloc(pVnode->inuse, reqSize);
+  if (pReq == NULL) {
+    // TODO: handle error
+  }
+
+  memcpy(pReq, pRequest, reqSize);
+
+  // Push the request to TQ so consumers can consume
+  tqPushMsg(pVnode->pTq, pReq, 0);
+
+  // Process the request
+  switch (reqType) {
+    case TSDB_MSG_TYPE_CREATE_TABLE:
+      code = metaCreateTable(pVnode->pMeta, NULL /* TODO */);
+      break;
+    case TSDB_MSG_TYPE_DROP_TABLE:
+      code = metaDropTable(pVnode->pMeta, 0 /* TODO */);
+      break;
+      /* TODO */
+    default:
+      break;
+  }
+
+  if (vnodeShouldCommit(pVnode)) {
+    if (vnodeAsyncCommit(pVnode) < 0) {
+      // TODO: handle error
+    }
+  }
+
+  return code;
+}
+
+/* ------------------------ STATIC METHODS ------------------------ */

source/dnode/vnode/impl/test/CMakeLists.txt

@@ -1,9 +1,7 @@
-# vnodeMemAllocatorTest
-add_executable(VMATest "")
-target_sources(VMATest
+# Vnode API test
+add_executable(vnodeApiTests "")
+target_sources(vnodeApiTests
     PRIVATE
-    "../src/vnodeMemAllocator.c"
-    "vnodeMemAllocatorTest.cpp"
+    "vnodeApiTests.cpp"
 )
-target_include_directories(VMATest PUBLIC "${CMAKE_CURRENT_SOURCE_DIR}/../inc")
-target_link_libraries(VMATest os gtest_main vnode)
+target_link_libraries(vnodeApiTests vnode gtest gtest_main)

source/dnode/vnode/impl/test/vnodeApiTests.cpp

@@ -0,0 +1,13 @@
+#include <gtest/gtest.h>
+#include <iostream>
+
+#include "vnode.h"
+
+TEST(vnodeApiTest, vnodeOpen_vnodeClose_test) {
+  // Create and open a vnode
+  SVnode *pVnode = vnodeOpen("vnode1", NULL);
+  ASSERT_NE(pVnode, nullptr);
+
+  // Close the vnode
+  vnodeClose(pVnode);
+}

source/dnode/vnode/impl/test/vnodeMemAllocatorTest.cpp

@@ -1,22 +0,0 @@
-#include <gtest/gtest.h>
-#include <iostream>
-
-#include "vnodeMemAllocator.h"
-
-TEST(VMATest, basic_create_and_destroy_test) {
-  SVnodeMemAllocator *vma = VMACreate(1024, 512, 64);
-  EXPECT_TRUE(vma != nullptr);
-  EXPECT_EQ(vma->full, false);
-  EXPECT_EQ(vma->ssize, 512);
-  EXPECT_EQ(vma->threshold, 64);
-  EXPECT_EQ(vma->inuse->tsize, 1024);
-  VMADestroy(vma);
-
-  vma = VMACreate(1024, 512, 1024);
-  EXPECT_TRUE(vma != nullptr);
-  VMADestroy(vma);
-
-  vma = VMACreate(1024, 512, 1025);
-  EXPECT_TRUE(vma == nullptr);
-  VMADestroy(vma);
-}

source/dnode/vnode/meta/inc/metaCache.h

@@ -0,0 +1,36 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef _TD_META_CACHE_H_
+#define _TD_META_CACHE_H_
+
+#include "rocksdb/c.h"
+
+#include "meta.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+typedef rocksdb_cache_t meta_cache_t;
+
+int  metaOpenCache(SMeta *pMeta);
+void metaCloseCache(SMeta *pMeta);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /*_TD_META_CACHE_H_*/

source/dnode/vnode/meta/inc/metaDB.h

@@ -0,0 +1,44 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef _TD_META_DB_H_
+#define _TD_META_DB_H_
+
+#include "rocksdb/c.h"
+
+#include "meta.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+typedef struct {
+  rocksdb_t *tbDb;      // uid -> tb obj
+  rocksdb_t *nameDb;    // name -> uid
+  rocksdb_t *tagDb;     // uid -> tag
+  rocksdb_t *schemaDb;  // uid+version -> schema
+  rocksdb_t *mapDb;     // suid -> uid_list
+} meta_db_t;
+
+int  metaOpenDB(SMeta *pMeta);
+void metaCloseDB(SMeta *pMeta);
+int  metaSaveTableToDB(SMeta *pMeta, const STbOptions *pTbOptions);
+int  metaRemoveTableFromDb(SMeta *pMeta, tb_uid_t uid);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /*_TD_META_DB_H_*/

source/dnode/vnode/meta/inc/metaDef.h

@@ -16,22 +16,26 @@
 #ifndef _TD_META_DEF_H_
 #define _TD_META_DEF_H_
 
-#include "metaUid.h"
-#include "tkv.h"
+#include "meta.h"
+#include "metaCache.h"
+#include "metaDB.h"
+#include "metaIdx.h"
+#include "metaOptions.h"
+#include "metaTbOptions.h"
+#include "metaTbTag.h"
+#include "metaTbUid.h"
 
 #ifdef __cplusplus
 extern "C" {
 #endif
 
 struct SMeta {
-  STableUidGenerator uidGenerator;
-
-  STkvDb* tableDb;   // uid->table obj
-  STkvDb* tbnameDb;  // tbname --> uid
-  STkvDb* schemaDb;  // uid+version --> schema
-  STkvDb* tagDb;     // uid --> tag
-  STkvDb* tagIdx;    // TODO: need to integrate lucene or our own
-  // STkvCache* metaCache; // TODO: add a global cache here
+  char*           path;     // path of current meta
+  SMetaOptions    options;  // meta option
+  meta_db_t*      pDB;      // raw data db
+  meta_index_t*   pIdx;     // tag index
+  meta_cache_t*   pCache;   // LRU cache
+  STbUidGenerator uidGnrt;  // meta table UID generator
 };
 
 #ifdef __cplusplus

source/dnode/vnode/meta/inc/metaIdx.h

@@ -0,0 +1,38 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef _TD_META_IDX_H_
+#define _TD_META_IDX_H_
+
+#include "rocksdb/c.h"
+
+#include "meta.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+typedef rocksdb_t meta_index_t;
+
+int  metaOpenIdx(SMeta *pMeta);
+void metaCloseIdx(SMeta *pMeta);
+int  metaSaveTableToIdx(SMeta *pMeta, const STbOptions *pTbOptions);
+int  metaRemoveTableFromIdx(SMeta *pMeta, tb_uid_t uid);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /*_TD_META_IDX_H_*/

source/dnode/vnode/meta/inc/metaOptions.h

@@ -0,0 +1,34 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef _TD_META_OPTIONS_H_
+#define _TD_META_OPTIONS_H_
+
+#include "meta.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+extern const SMetaOptions defaultMetaOptions;
+
+int  metaValidateOptions(const SMetaOptions *);
+void metaOptionsCopy(SMetaOptions *pDest, const SMetaOptions *pSrc);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /*_TD_META_OPTIONS_H_*/

source/dnode/vnode/meta/inc/metaQuery.h

@@ -0,0 +1,27 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef _TD_META_QUERY_H_
+#define _TD_META_QUERY_H_
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /*_TD_META_QUERY_H_*/

source/dnode/vnode/meta/inc/metaTbOptions.h

@@ -0,0 +1,32 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef _TD_META_TABLE_OPTIONS_H_
+#define _TD_META_TABLE_OPTIONS_H_
+
+#include "meta.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+int    metaValidateTbOptions(SMeta *pMeta, const STbOptions *);
+size_t metaEncodeTbObjFromTbOptions(const STbOptions *, void *pBuf, size_t bsize);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /*_TD_META_TABLE_OPTIONS_H_*/

source/dnode/vnode/meta/inc/metaTbTag.h

@@ -0,0 +1,27 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef _TD_META_TB_TAG_H_
+#define _TD_META_TB_TAG_H_
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /*_TD_META_TB_TAG_H_*/

source/dnode/vnode/meta/inc/metaTbUid.h

@@ -23,20 +23,17 @@ extern "C" {
 #endif
 
 /* ------------------------ APIS EXPOSED ------------------------ */
-typedef struct STableUidGenerator STableUidGenerator;
+typedef struct STbUidGenerator {
+  tb_uid_t nextUid;
+} STbUidGenerator;
+
+// STableUidGenerator
+int  metaOpenUidGnrt(SMeta *pMeta);
+void metaCloseUidGnrt(SMeta *pMeta);
 
 // tb_uid_t
 #define IVLD_TB_UID 0
-tb_uid_t generateUid(STableUidGenerator *);
-
-// STableUidGenerator
-void tableUidGeneratorInit(STableUidGenerator *, tb_uid_t suid);
-#define tableUidGeneratorClear(ug)
-
-/* ------------------------ FOR TEST AND COMPILE ONLY ------------------------ */
-struct STableUidGenerator {
-  tb_uid_t nextUid;
-};
+tb_uid_t metaGenerateUid(SMeta *pMeta);
 
 #ifdef __cplusplus
 }

source/dnode/vnode/meta/src/metaCache.c

@@ -0,0 +1,37 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "meta.h"
+#include "metaDef.h"
+
+int metaOpenCache(SMeta *pMeta) {
+  // TODO
+  if (pMeta->options.lruCacheSize) {
+    pMeta->pCache = rocksdb_cache_create_lru(pMeta->options.lruCacheSize);
+    if (pMeta->pCache == NULL) {
+      // TODO: handle error
+      return -1;
+    }
+  }
+
+  return 0;
+}
+
+void metaCloseCache(SMeta *pMeta) {
+  if (pMeta->pCache) {
+    rocksdb_cache_destroy(pMeta->pCache);
+    pMeta->pCache = NULL;
+  }
+}

source/dnode/vnode/meta/src/metaCommit.c

@@ -0,0 +1,21 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "meta.h"
+
+int metaCommit(SMeta *pMeta) {
+  // TODO
+  return 0;
+}

source/dnode/vnode/meta/src/metaDB.c

@@ -0,0 +1,199 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "metaDef.h"
+
+static void metaSaveSchemaDB(SMeta *pMeta, tb_uid_t uid, STSchema *pSchema);
+static void metaGetSchemaDBKey(char key[], tb_uid_t uid, int sversion);
+static int  metaSaveMapDB(SMeta *pMeta, tb_uid_t suid, tb_uid_t uid);
+
+#define SCHEMA_KEY_LEN (sizeof(tb_uid_t) + sizeof(int))
+
+#define META_OPEN_DB_IMPL(pDB, options, dir, err) \
+  do {                                            \
+    pDB = rocksdb_open(options, dir, &err);       \
+    if (pDB == NULL) {                            \
+      metaCloseDB(pMeta);                         \
+      rocksdb_options_destroy(options);           \
+      return -1;                                  \
+    }                                             \
+  } while (0)
+
+int metaOpenDB(SMeta *pMeta) {
+  char               dir[128];
+  char *             err = NULL;
+  rocksdb_options_t *options = rocksdb_options_create();
+
+  if (pMeta->pCache) {
+    rocksdb_options_set_row_cache(options, pMeta->pCache);
+  }
+  rocksdb_options_set_create_if_missing(options, 1);
+
+  pMeta->pDB = (meta_db_t *)calloc(1, sizeof(*(pMeta->pDB)));
+  if (pMeta->pDB == NULL) {
+    // TODO: handle error
+    rocksdb_options_destroy(options);
+    return -1;
+  }
+
+  // tbDb
+  sprintf(dir, "%s/tb_db", pMeta->path);
+  META_OPEN_DB_IMPL(pMeta->pDB->tbDb, options, dir, err);
+
+  // nameDb
+  sprintf(dir, "%s/name_db", pMeta->path);
+  META_OPEN_DB_IMPL(pMeta->pDB->nameDb, options, dir, err);
+
+  // tagDb
+  sprintf(dir, "%s/tag_db", pMeta->path);
+  META_OPEN_DB_IMPL(pMeta->pDB->tagDb, options, dir, err);
+
+  // schemaDb
+  sprintf(dir, "%s/schema_db", pMeta->path);
+  META_OPEN_DB_IMPL(pMeta->pDB->schemaDb, options, dir, err);
+
+  // mapDb
+  sprintf(dir, "%s/map_db", pMeta->path);
+  META_OPEN_DB_IMPL(pMeta->pDB->mapDb, options, dir, err);
+
+  rocksdb_options_destroy(options);
+  return 0;
+}
+
+#define META_CLOSE_DB_IMPL(pDB) \
+  do {                          \
+    if (pDB) {                  \
+      rocksdb_close(pDB);       \
+      pDB = NULL;               \
+    }                           \
+  } while (0)
+
+void metaCloseDB(SMeta *pMeta) {
+  if (pMeta->pDB) {
+    META_CLOSE_DB_IMPL(pMeta->pDB->mapDb);
+    META_CLOSE_DB_IMPL(pMeta->pDB->schemaDb);
+    META_CLOSE_DB_IMPL(pMeta->pDB->tagDb);
+    META_CLOSE_DB_IMPL(pMeta->pDB->nameDb);
+    META_CLOSE_DB_IMPL(pMeta->pDB->tbDb);
+    free(pMeta->pDB);
+    pMeta->pDB = NULL;
+  }
+}
+
+int metaSaveTableToDB(SMeta *pMeta, const STbOptions *pTbOptions) {
+  tb_uid_t uid;
+  char *   err = NULL;
+  size_t   size;
+  char     pBuf[1024];  // TODO
+
+  rocksdb_writeoptions_t *wopt = rocksdb_writeoptions_create();
+
+  // Generate a uid for child and normal table
+  if (pTbOptions->type == META_SUPER_TABLE) {
+    uid = pTbOptions->stbOptions.uid;
+  } else {
+    uid = metaGenerateUid(pMeta);
+  }
+
+  // Save tbname -> uid to tbnameDB
+  rocksdb_put(pMeta->pDB->nameDb, wopt, pTbOptions->name, strlen(pTbOptions->name), (char *)(&uid), sizeof(uid), &err);
+
+  // Save uid -> tb_obj to tbDB
+  size = metaEncodeTbObjFromTbOptions(pTbOptions, pBuf, 1024);
+  rocksdb_put(pMeta->pDB->tbDb, wopt, (char *)(&uid), sizeof(uid), pBuf, size, &err);
+
+  switch (pTbOptions->type) {
+    case META_NORMAL_TABLE:
+      // save schemaDB
+      metaSaveSchemaDB(pMeta, uid, pTbOptions->ntbOptions.pSchame);
+      break;
+    case META_SUPER_TABLE:
+      // save schemaDB
+      metaSaveSchemaDB(pMeta, uid, pTbOptions->stbOptions.pSchema);
+
+      // save mapDB (really need?)
+      rocksdb_put(pMeta->pDB->mapDb, wopt, (char *)(&uid), sizeof(uid), "", 0, &err);
+      break;
+    case META_CHILD_TABLE:
+      // save tagDB
+      rocksdb_put(pMeta->pDB->tagDb, wopt, (char *)(&uid), sizeof(uid), pTbOptions->ctbOptions.tags,
+                  kvRowLen(pTbOptions->ctbOptions.tags), &err);
+
+      // save mapDB
+      metaSaveMapDB(pMeta, pTbOptions->ctbOptions.suid, uid);
+      break;
+    default:
+      ASSERT(0);
+  }
+
+  rocksdb_writeoptions_destroy(wopt);
+
+  return 0;
+}
+
+int metaRemoveTableFromDb(SMeta *pMeta, tb_uid_t uid) {
+  /* TODO */
+  return 0;
+}
+
+/* ------------------------ STATIC METHODS ------------------------ */
+static void metaSaveSchemaDB(SMeta *pMeta, tb_uid_t uid, STSchema *pSchema) {
+  char   key[64];
+  char   pBuf[1024];
+  char * ppBuf = pBuf;
+  size_t vsize;
+  char * err = NULL;
+
+  rocksdb_writeoptions_t *wopt = rocksdb_writeoptions_create();
+
+  metaGetSchemaDBKey(key, uid, schemaVersion(pSchema));
+  vsize = tdEncodeSchema((void **)(&ppBuf), pSchema);
+  rocksdb_put(pMeta->pDB->schemaDb, wopt, key, SCHEMA_KEY_LEN, pBuf, vsize, &err);
+
+  rocksdb_writeoptions_destroy(wopt);
+}
+
+static void metaGetSchemaDBKey(char *key, tb_uid_t uid, int sversion) {
+  *(tb_uid_t *)key = uid;
+  *(int *)POINTER_SHIFT(key, sizeof(tb_uid_t)) = sversion;
+}
+
+static int metaSaveMapDB(SMeta *pMeta, tb_uid_t suid, tb_uid_t uid) {
+  size_t vlen;
+  char * val;
+  char * err = NULL;
+
+  rocksdb_readoptions_t *ropt = rocksdb_readoptions_create();
+  val = rocksdb_get(pMeta->pDB->mapDb, ropt, (char *)(&suid), sizeof(suid), &vlen, &err);
+  rocksdb_readoptions_destroy(ropt);
+
+  void *nval = malloc(vlen + sizeof(uid));
+  if (nval == NULL) {
+    return -1;
+  }
+
+  if (vlen) {
+    memcpy(nval, val, vlen);
+  }
+  memcpy(POINTER_SHIFT(nval, vlen), (void *)(&uid), sizeof(uid));
+
+  rocksdb_writeoptions_t *wopt = rocksdb_writeoptions_create();
+
+  rocksdb_put(pMeta->pDB->mapDb, wopt, (char *)(&suid), sizeof(suid), nval, vlen + sizeof(uid), &err);
+
+  rocksdb_writeoptions_destroy(wopt);
+
+  return 0;
+}

source/dnode/vnode/meta/src/metaIdx.c

@@ -0,0 +1,53 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "metaDef.h"
+
+int metaOpenIdx(SMeta *pMeta) {
+  char               idxDir[128];  // TODO
+  char *             err = NULL;
+  rocksdb_options_t *options = rocksdb_options_create();
+
+  // TODO
+  sprintf(idxDir, "%s/index", pMeta->path);
+
+  if (pMeta->pCache) {
+    rocksdb_options_set_row_cache(options, pMeta->pCache);
+  }
+  rocksdb_options_set_create_if_missing(options, 1);
+
+  pMeta->pIdx = rocksdb_open(options, idxDir, &err);
+  if (pMeta->pIdx == NULL) {
+    // TODO: handle error
+    rocksdb_options_destroy(options);
+    return -1;
+  }
+
+  rocksdb_options_destroy(options);
+
+  return 0;
+}
+
+void metaCloseIdx(SMeta *pMeta) { /* TODO */
+  if (pMeta->pIdx) {
+    rocksdb_close(pMeta->pIdx);
+    pMeta->pIdx = NULL;
+  }
+}
+
+int metaSaveTableToIdx(SMeta *pMeta, const STbOptions *pTbOptions) {
+  // TODO
+  return 0;
+}

source/dnode/vnode/meta/src/metaMain.c

@@ -13,75 +13,134 @@
  * along with this program. If not, see <http://www.gnu.org/licenses/>.
  */
 
-#include "meta.h"
-#include "metaDef.h"
 #include "tcoding.h"
 
-static int metaCreateSuperTable(SMeta *pMeta, const char *tbname, const SSuperTableOpts *pSuperTableOpts);
-static int metaCreateChildTable(SMeta *pMeta, const char *tbname, const SChildTableOpts *pChildTableOpts);
-static int metaCreateNormalTable(SMeta *pMeta, const char *tbname, const SNormalTableOpts *pNormalTableOpts);
+#include "meta.h"
+#include "metaDB.h"
+#include "metaDef.h"
+#include "metaOptions.h"
 
-SMeta *metaOpen(SMetaOpts *pMetaOpts) {
+static SMeta *metaNew(const char *path, const SMetaOptions *pMetaOptions);
+static void   metaFree(SMeta *pMeta);
+static int    metaOpenImpl(SMeta *pMeta);
+static void   metaCloseImpl(SMeta *pMeta);
+
+SMeta *metaOpen(const char *path, const SMetaOptions *pMetaOptions) {
   SMeta *pMeta = NULL;
 
+  // Set default options
+  if (pMetaOptions == NULL) {
+    pMetaOptions = &defaultMetaOptions;
+  }
+
+  // Validate the options
+  if (metaValidateOptions(pMetaOptions) < 0) {
+    // TODO: deal with error
+    return NULL;
+  }
+
+  // Allocate handle
+  pMeta = metaNew(path, pMetaOptions);
+  if (pMeta == NULL) {
+    // TODO: handle error
+    return NULL;
+  }
+
+  // Create META path (TODO)
+  taosMkDir(path);
+
+  // Open meta
+  if (metaOpenImpl(pMeta) < 0) {
+    metaFree(pMeta);
+    return NULL;
+  }
+
+  return pMeta;
+}
+
+void metaClose(SMeta *pMeta) {
+  if (pMeta) {
+    metaCloseImpl(pMeta);
+    metaFree(pMeta);
+  }
+}
+
+void metaRemove(const char *path) { taosRemoveDir(path); }
+
+/* ------------------------ STATIC METHODS ------------------------ */
+static SMeta *metaNew(const char *path, const SMetaOptions *pMetaOptions) {
+  SMeta *pMeta;
+  size_t psize = strlen(path);
+
   pMeta = (SMeta *)calloc(1, sizeof(*pMeta));
   if (pMeta == NULL) {
     return NULL;
   }
 
-  // TODO: check if file exists and handle the error
-  taosMkDir("meta");
+  pMeta->path = strdup(path);
+  if (pMeta->path == NULL) {
+    metaFree(pMeta);
+    return NULL;
+  }
 
-  // Open tableDb
-  STkvOpts *tableDbOpts = tkvOptsCreate();
-  tkvOptsSetCreateIfMissing(tableDbOpts, 1);
-  pMeta->tableDb = tkvOpen(tableDbOpts, "meta/table_db");
-  tkvOptsDestroy(tableDbOpts);
+  metaOptionsCopy(&(pMeta->options), pMetaOptions);
 
-  // Open tbnameDb
-  STkvOpts *tbnameDbOpts = tkvOptsCreate();
-  tkvOptsSetCreateIfMissing(tbnameDbOpts, 1);
-  pMeta->tbnameDb = tkvOpen(tbnameDbOpts, "meta/tbname_db");
-  tkvOptsDestroy(tbnameDbOpts);
-
-  // Open schemaDb
-  STkvOpts *schemaDbOpts = tkvOptsCreate();
-  tkvOptsSetCreateIfMissing(schemaDbOpts, 1);
-  pMeta->schemaDb = tkvOpen(schemaDbOpts, "meta/schema_db");
-  tkvOptsDestroy(schemaDbOpts);
-
-  // Open tagDb
-  STkvOpts *tagDbOpts = tkvOptsCreate();
-  tkvOptsSetCreateIfMissing(tagDbOpts, 1);
-  pMeta->tagDb = tkvOpen(tagDbOpts, "meta/tag_db");
-  tkvOptsDestroy(tagDbOpts);
-
-  // Open tagIdx
-  STkvOpts *tagIdxDbOpts = tkvOptsCreate();
-  tkvOptsSetCreateIfMissing(tagIdxDbOpts, 1);
-  pMeta->tagIdx = tkvOpen(tagIdxDbOpts, "meta/tag_idx_db");
-  tkvOptsDestroy(tagIdxDbOpts);
-
-  // TODO: need to figure out how to persist the START UID
-  tableUidGeneratorInit(&(pMeta->uidGenerator), IVLD_TB_UID);
   return pMeta;
-}
+};
 
-void metaClose(SMeta *pMeta) {
+static void metaFree(SMeta *pMeta) {
   if (pMeta) {
-    tableUidGeneratorClear(&pMeta->uidGenerator);
-
-    tkvClose(pMeta->tagIdx);
-    tkvClose(pMeta->tagDb);
-    tkvClose(pMeta->schemaDb);
-    tkvClose(pMeta->tbnameDb);
-    tkvClose(pMeta->tableDb);
-
+    tfree(pMeta->path);
     free(pMeta);
   }
 }
 
-int metaCreateTable(SMeta *pMeta, const STableOpts *pTableOpts) {
+static int metaOpenImpl(SMeta *pMeta) {
+  // Open meta cache
+  if (metaOpenCache(pMeta) < 0) {
+    // TODO: handle error
+    metaCloseImpl(pMeta);
+    return -1;
+  }
+
+  // Open meta db
+  if (metaOpenDB(pMeta) < 0) {
+    // TODO: handle error
+    metaCloseImpl(pMeta);
+    return -1;
+  }
+
+  // Open meta index
+  if (metaOpenIdx(pMeta) < 0) {
+    // TODO: handle error
+    metaCloseImpl(pMeta);
+    return -1;
+  }
+
+  // Open meta table uid generator
+  if (metaOpenUidGnrt(pMeta) < 0) {
+    // TODO: handle error
+    metaCloseImpl(pMeta);
+    return -1;
+  }
+
+  return 0;
+}
+
+static void metaCloseImpl(SMeta *pMeta) {
+  metaCloseUidGnrt(pMeta);
+  metaCloseIdx(pMeta);
+  metaCloseDB(pMeta);
+  metaCloseCache(pMeta);
+}
+
+// OLD -------------------------------------------------------------------
+#if 0
+static int    metaCreateSuperTable(SMeta *pMeta, const char *tbname, const SSuperTableOpts *pSuperTableOpts);
+static int    metaCreateChildTable(SMeta *pMeta, const char *tbname, const SChildTableOpts *pChildTableOpts);
+static int    metaCreateNormalTable(SMeta *pMeta, const char *tbname, const SNormalTableOpts *pNormalTableOpts);
+
+int metaCreateTable(SMeta *pMeta, const STableOptions *pTableOpts) {
   size_t vallen;
   char * pUid;
 
@@ -107,7 +166,6 @@ int metaCreateTable(SMeta *pMeta, const STableOpts *pTableOpts) {
   return 0;
 }
 
-/* ------------------------ STATIC METHODS ------------------------ */
 static int metaCreateSuperTable(SMeta *pMeta, const char *tbname, const SSuperTableOpts *pSuperTableOpts) {
   size_t vallen;
   size_t keylen;
@@ -213,13 +271,13 @@ static int metaCreateNormalTable(SMeta *pMeta, const char *tbname, const SNormal
   return 0;
 }
 
-void metaNormalTableOptsInit(STableOpts *pTableOpts, const char *name, const STSchema *pSchema) {
+void metaNormalTableOptsInit(STableOptions *pTableOpts, const char *name, const STSchema *pSchema) {
   pTableOpts->type = META_NORMAL_TABLE;
   pTableOpts->name = strdup(name);
   pTableOpts->normalOpts.pSchema = tdDupSchema(pSchema);
 }
 
-void metaSuperTableOptsInit(STableOpts *pTableOpts, const char *name, tb_uid_t uid, const STSchema *pSchema,
+void metaSuperTableOptsInit(STableOptions *pTableOpts, const char *name, tb_uid_t uid, const STSchema *pSchema,
                             const STSchema *pTagSchema) {
   pTableOpts->type = META_SUPER_TABLE;
   pTableOpts->name = strdup(name);
@@ -228,14 +286,14 @@ void metaSuperTableOptsInit(STableOpts *pTableOpts, const char *name, tb_uid_t u
   pTableOpts->superOpts.pTagSchema = tdDupSchema(pTagSchema);
 }
 
-void metaChildTableOptsInit(STableOpts *pTableOpts, const char *name, tb_uid_t suid, const SKVRow tags) {
+void metaChildTableOptsInit(STableOptions *pTableOpts, const char *name, tb_uid_t suid, const SKVRow tags) {
   pTableOpts->type = META_CHILD_TABLE;
   pTableOpts->name = strdup(name);
   pTableOpts->childOpts.suid = suid;
   pTableOpts->childOpts.tags = tdKVRowDup(tags);
 }
 
-void metaTableOptsClear(STableOpts *pTableOpts) {
+void metaTableOptsClear(STableOptions *pTableOpts) {
   switch (pTableOpts->type) {
     case META_NORMAL_TABLE:
       tfree(pTableOpts->name);
@@ -258,3 +316,5 @@ void metaTableOptsClear(STableOpts *pTableOpts) {
 }
 
 void metaDestroy(const char *path) { taosRemoveDir(path); }
+
+#endif

source/dnode/vnode/meta/src/metaOptions.c

@@ -0,0 +1,34 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "metaDef.h"
+
+const SMetaOptions defaultMetaOptions = {.lruCacheSize = 0};
+
+/* ------------------------ EXPOSED METHODS ------------------------ */
+void metaOptionsInit(SMetaOptions *pMetaOptions) { metaOptionsCopy(pMetaOptions, &defaultMetaOptions); }
+
+void metaOptionsClear(SMetaOptions *pMetaOptions) {
+  // TODO
+}
+
+int metaValidateOptions(const SMetaOptions *pMetaOptions) {
+  // TODO
+  return 0;
+}
+
+void metaOptionsCopy(SMetaOptions *pDest, const SMetaOptions *pSrc) { memcpy(pDest, pSrc, sizeof(*pSrc)); }
+
+/* ------------------------ STATIC METHODS ------------------------ */

source/dnode/vnode/meta/src/metaQuery.c

@@ -0,0 +1,14 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */

source/dnode/vnode/meta/src/metaTable.c

@@ -0,0 +1,52 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "metaDef.h"
+
+int metaCreateTable(SMeta *pMeta, const STbOptions *pTbOptions) {
+  // Validate the tbOptions
+  if (metaValidateTbOptions(pMeta, pTbOptions) < 0) {
+    // TODO: handle error
+    return -1;
+  }
+
+  // TODO: add atomicity
+
+  if (metaSaveTableToDB(pMeta, pTbOptions) < 0) {
+    // TODO: handle error
+    return -1;
+  }
+
+  if (metaSaveTableToIdx(pMeta, pTbOptions) < 0) {
+    // TODO: handle error
+    return -1;
+  }
+
+  return 0;
+}
+
+int metaDropTable(SMeta *pMeta, tb_uid_t uid) {
+  if (metaRemoveTableFromIdx(pMeta, uid) < 0) {
+    // TODO: handle error
+    return -1;
+  }
+
+  if (metaRemoveTableFromIdx(pMeta, uid) < 0) {
+    // TODO
+    return -1;
+  }
+
+  return 0;
+}

source/dnode/vnode/meta/src/metaTbOptions.c

@@ -0,0 +1,49 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "metaDef.h"
+#include "tcoding.h"
+
+int metaValidateTbOptions(SMeta *pMeta, const STbOptions *pTbOptions) {
+  // TODO
+  return 0;
+}
+
+size_t metaEncodeTbObjFromTbOptions(const STbOptions *pTbOptions, void *pBuf, size_t bsize) {
+  void **ppBuf = &pBuf;
+  int    tlen = 0;
+
+  tlen += taosEncodeFixedU8(ppBuf, pTbOptions->type);
+  tlen += taosEncodeString(ppBuf, pTbOptions->name);
+  tlen += taosEncodeFixedU32(ppBuf, pTbOptions->ttl);
+
+  switch (pTbOptions->type) {
+    case META_SUPER_TABLE:
+      tlen += taosEncodeFixedU64(ppBuf, pTbOptions->stbOptions.uid);
+      tlen += tdEncodeSchema(ppBuf, pTbOptions->stbOptions.pTagSchema);
+      // TODO: encode schema version array
+      break;
+    case META_CHILD_TABLE:
+      tlen += taosEncodeFixedU64(ppBuf, pTbOptions->ctbOptions.suid);
+      break;
+    case META_NORMAL_TABLE:
+      // TODO: encode schema version array
+      break;
+    default:
+      break;
+  }
+
+  return tlen;
+}

source/dnode/vnode/meta/src/metaTbTag.c

@@ -0,0 +1,14 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */

source/dnode/vnode/meta/src/metaTbUid.c

@@ -13,14 +13,18 @@
  * along with this program. If not, see <http://www.gnu.org/licenses/>.
  */
 
-#include "metaUid.h"
+#include "meta.h"
+#include "metaDef.h"
 
-tb_uid_t generateUid(STableUidGenerator *pGen) {
-  // Generate a new table UID
-  return ++(pGen->nextUid);
+int metaOpenUidGnrt(SMeta *pMeta) {
+  // Init a generator
+  pMeta->uidGnrt.nextUid = IVLD_TB_UID;
+  return 0;
 }
 
-void tableUidGeneratorInit(STableUidGenerator *pGen, tb_uid_t suid) {
-  // Init a generator
-  pGen->nextUid = suid;
+void metaCloseUidGnrt(SMeta *pMeta) { /* TODO */ }
+
+tb_uid_t metaGenerateUid(SMeta *pMeta) {
+  // Generate a new table UID
+  return ++(pMeta->uidGnrt.nextUid);
 }

source/dnode/vnode/meta/test/CMakeLists.txt

@@ -1,24 +1,24 @@
-add_executable(metaTest "")
-target_sources(metaTest
-    PRIVATE
-    "../src/metaMain.c"
-    "../src/metaUid.c"
-    "metaTests.cpp"
-)
-target_include_directories(metaTest 
-    PUBLIC
-    "${CMAKE_SOURCE_DIR}/include/server/vnode/meta"
-    "${CMAKE_CURRENT_SOURCE_DIR}/../inc"
-)
-target_link_libraries(metaTest
-    os
-    util
-    common
-    gtest_main
-    tkv
-)
-enable_testing()
-add_test(
-    NAME meta_test
-    COMMAND metaTest
-)
+# add_executable(metaTest "")
+# target_sources(metaTest
+#     PRIVATE
+#     "../src/metaMain.c"
+#     "../src/metaUid.c"
+#     "metaTests.cpp"
+# )
+# target_include_directories(metaTest 
+#     PUBLIC
+#     "${CMAKE_SOURCE_DIR}/include/server/vnode/meta"
+#     "${CMAKE_CURRENT_SOURCE_DIR}/../inc"
+# )
+# target_link_libraries(metaTest
+#     os
+#     util
+#     common
+#     gtest_main
+#     tkv
+# )
+# enable_testing()
+# add_test(
+#     NAME meta_test
+#     COMMAND metaTest
+# )

source/dnode/vnode/meta/test/metaAPITest.cpp

@@ -1,3 +1,4 @@
+#if 0
 #include <gtest/gtest.h>
 #include <string.h>
 #include <iostream>
@@ -38,7 +39,7 @@ static SKVRow metaGetSimpleTags() {
 
 TEST(MetaTest, DISABLED_meta_create_1m_normal_tables_test) {
   // Open Meta
-  SMeta *meta = metaOpen(NULL);
+  SMeta *meta = metaOpen(NULL, NULL);
   std::cout << "Meta is opened!" << std::endl;
 
   // Create 1000000 normal tables
@@ -100,4 +101,5 @@ TEST(MetaTest, meta_create_1m_child_tables_test) {
   // Destroy Meta
   metaDestroy("meta");
   std::cout << "Meta is destroyed!" << std::endl;
-}
+}
+#endif

source/dnode/vnode/tq/inc/tqMetaStore.h

@@ -24,18 +24,18 @@
 extern "C" {
 #endif
 
-#define TQ_BUCKET_SIZE 0xFF
+#define TQ_BUCKET_MASK 0xFF
+#define TQ_BUCKET_SIZE 256
+
 #define TQ_PAGE_SIZE 4096
 //key + offset + size
-#define TQ_IDX_ENTRY_SIZE 24
-
-inline static int TqMaxEntryOnePage() { //170
-  return TQ_PAGE_SIZE / TQ_IDX_ENTRY_SIZE;
-}
-
-inline static int TqEmptyTail() { //16
-  return TQ_PAGE_SIZE - TqMaxEntryOnePage();
-}
+#define TQ_IDX_SIZE 24
+//4096 / 24
+#define TQ_MAX_IDX_ONE_PAGE 170
+//24 * 170
+#define TQ_IDX_PAGE_BODY_SIZE 4080
+//4096 - 4080
+#define TQ_IDX_PAGE_HEAD_SIZE 16
 
 #define TQ_ACTION_CONST      0
 #define TQ_ACTION_INUSE      1
@@ -92,19 +92,17 @@ int32_t       tqStoreClose(TqMetaStore*);
 //int32_t       tqStoreDelete(TqMetaStore*);
 //int32_t       TqStoreCommitAll(TqMetaStore*);
 int32_t       tqStorePersist(TqMetaStore*);
+//clean deleted idx and data from persistent file
+int32_t       tqStoreCompact(TqMetaStore*);
 
 void*   tqHandleGet(TqMetaStore*, int64_t key);
 int32_t tqHandleMovePut(TqMetaStore*, int64_t key, void* value);
 int32_t tqHandleCopyPut(TqMetaStore*, int64_t key, void* value, size_t vsize);
-//do commit
-int32_t tqHandleCommit(TqMetaStore*, int64_t key);
-//delete uncommitted
-int32_t tqHandleAbort(TqMetaStore*, int64_t key);
 //delete committed kv pair
 //notice that a delete action still needs to be committed
 int32_t tqHandleDel(TqMetaStore*, int64_t key);
-//delete both committed and uncommitted
-int32_t tqHandleClear(TqMetaStore*, int64_t key);
+int32_t tqHandleCommit(TqMetaStore*, int64_t key);
+int32_t tqHandleAbort(TqMetaStore*, int64_t key);
 
 #ifdef __cplusplus
 }

source/dnode/vnode/tq/src/tq.c

@@ -22,6 +22,9 @@
 //
 //handle management message
 //
+
+int tqGetgHandleSSize(const TqGroupHandle *gHandle);
+
 static int tqProtoCheck(TmqMsgHead *pMsg) {
   return pMsg->protoVer == 0;
 }

source/dnode/vnode/tq/src/tqMetaStore.c

@@ -27,19 +27,46 @@ static int32_t tqHandlePutCommitted(TqMetaStore*, int64_t key, void* value);
 static void*   tqHandleGetUncommitted(TqMetaStore*, int64_t key);
 
 static inline void tqLinkUnpersist(TqMetaStore *pMeta, TqMetaList* pNode) {
-    if(pNode->unpersistNext == NULL) {
-      pNode->unpersistNext = pMeta->unpersistHead->unpersistNext;
-      pNode->unpersistPrev = pMeta->unpersistHead;
-      pMeta->unpersistHead->unpersistNext->unpersistPrev = pNode;
-      pMeta->unpersistHead->unpersistNext = pNode;
-    }
+  if(pNode->unpersistNext == NULL) {
+    pNode->unpersistNext = pMeta->unpersistHead->unpersistNext;
+    pNode->unpersistPrev = pMeta->unpersistHead;
+    pMeta->unpersistHead->unpersistNext->unpersistPrev = pNode;
+    pMeta->unpersistHead->unpersistNext = pNode;
+  }
 }
 
+static inline int tqSeekLastPage(int fd) {
+  int offset = lseek(fd, 0, SEEK_END);
+  int pageNo = offset / TQ_PAGE_SIZE;
+  int curPageOffset = pageNo * TQ_PAGE_SIZE;
+  return lseek(fd, curPageOffset, SEEK_SET);
+}
 
-typedef struct TqMetaPageBuf {
-  int16_t offset;
-  char buffer[TQ_PAGE_SIZE];
-} TqMetaPageBuf;
+//TODO: the struct is tightly coupled with index entry
+typedef struct TqIdxPageHead {
+  int16_t writeOffset;
+  int8_t  unused[14];
+} TqIdxPageHead;
+
+typedef struct TqIdxPageBuf {
+  TqIdxPageHead head;
+  char buffer[TQ_IDX_PAGE_BODY_SIZE];
+} TqIdxPageBuf;
+
+static inline int tqReadLastPage(int fd, TqIdxPageBuf* pBuf) {
+  int offset = tqSeekLastPage(fd);
+  int nBytes;
+  if((nBytes = read(fd, pBuf, TQ_PAGE_SIZE)) == -1) {
+    return -1;
+  }
+  if(nBytes == 0) {
+    memset(pBuf, 0, TQ_PAGE_SIZE);
+    pBuf->head.writeOffset = TQ_IDX_PAGE_HEAD_SIZE;
+  }
+  ASSERT(nBytes == 0 || nBytes == pBuf->head.writeOffset);
+
+  return lseek(fd, offset, SEEK_SET);
+}
 
 TqMetaStore* tqStoreOpen(const char* path,
     int serializer(const void* pObj, TqSerializedHead** ppHead),
@@ -103,27 +130,30 @@ TqMetaStore* tqStoreOpen(const char* path,
   pMeta->deleter = deleter;
 
   //read idx file and load into memory
-  char idxBuf[TQ_PAGE_SIZE];
+  TqIdxPageBuf idxBuf;
   TqSerializedHead* serializedObj = malloc(TQ_PAGE_SIZE);
   if(serializedObj == NULL) {
     //TODO:memory insufficient
   }
   int idxRead;
   int allocated = TQ_PAGE_SIZE;
-  while((idxRead = read(idxFd, idxBuf, TQ_PAGE_SIZE))) {
+  bool readEnd = false;
+  while((idxRead = read(idxFd, &idxBuf, TQ_PAGE_SIZE))) {
     if(idxRead == -1) {
       //TODO: handle error
       ASSERT(false);
     }
+    ASSERT(idxBuf.head.writeOffset == idxRead);
     //loop read every entry
-    for(int i = 0; i < idxRead; i += TQ_IDX_ENTRY_SIZE) {
+    for(int i = 0; i < idxBuf.head.writeOffset - TQ_IDX_PAGE_HEAD_SIZE; i += TQ_IDX_SIZE) {
       TqMetaList *pNode = malloc(sizeof(TqMetaList));
       if(pNode == NULL) {
         //TODO: free memory and return error
       }
       memset(pNode, 0, sizeof(TqMetaList));
-      memcpy(&pNode->handle, &idxBuf[i], TQ_IDX_ENTRY_SIZE);
-      lseek(fileFd, pNode->handle.offset, SEEK_CUR);
+      memcpy(&pNode->handle, &idxBuf.buffer[i], TQ_IDX_SIZE);
+
+      lseek(fileFd, pNode->handle.offset, SEEK_SET);
       if(allocated < pNode->handle.serializedSize) {
         void *ptr = realloc(serializedObj, pNode->handle.serializedSize);
         if(ptr == NULL) {
@@ -154,9 +184,9 @@ TqMetaStore* tqStoreOpen(const char* path,
         } else {
           pNode->handle.valueInUse = TQ_DELETE_TOKEN;
         }
-        serializedObj = POINTER_SHIFT(serializedObj, serializedObj->ssize);
-        if(serializedObj->ssize != sizeof(TqSerializedHead)) {
-          pMeta->deserializer(serializedObj, &pNode->handle.valueInTxn);
+        TqSerializedHead* ptr = POINTER_SHIFT(serializedObj, serializedObj->ssize);
+        if(ptr->ssize != sizeof(TqSerializedHead)) {
+          pMeta->deserializer(ptr, &pNode->handle.valueInTxn);
         } else {
           pNode->handle.valueInTxn = TQ_DELETE_TOKEN;
         }
@@ -165,7 +195,7 @@ TqMetaStore* tqStoreOpen(const char* path,
       }
 
       //put into list
-      int bucketKey = pNode->handle.key & TQ_BUCKET_SIZE;
+      int bucketKey = pNode->handle.key & TQ_BUCKET_MASK;
       TqMetaList* pBucketNode = pMeta->bucket[bucketKey];
       if(pBucketNode == NULL) {
         pMeta->bucket[bucketKey] = pNode;
@@ -174,15 +204,18 @@ TqMetaStore* tqStoreOpen(const char* path,
         pMeta->bucket[bucketKey] = pNode;
       } else {
         while(pBucketNode->next &&
-            pBucketNode->next->handle.key == pNode->handle.key) {
+            pBucketNode->next->handle.key != pNode->handle.key) {
           pBucketNode = pBucketNode->next; 
         }
         if(pBucketNode->next) {
           ASSERT(pBucketNode->next->handle.key == pNode->handle.key);
-          TqMetaList *pNodeTmp = pBucketNode->next;
-          pBucketNode->next = pNodeTmp->next;
-          pBucketNode = pNodeTmp;
+          TqMetaList *pNodeFound = pBucketNode->next;
+          pNode->next = pNodeFound->next;
+          pBucketNode->next = pNode;
+          pBucketNode = pNodeFound;
         } else {
+          pNode->next = pMeta->bucket[bucketKey];
+          pMeta->bucket[bucketKey] = pNode;
           pBucketNode = NULL;
         }
       }
@@ -264,8 +297,8 @@ int32_t tqStoreDelete(TqMetaStore* pMeta) {
 
 //TODO: wrap in tfile
 int32_t tqStorePersist(TqMetaStore* pMeta) {
-  char writeBuf[TQ_PAGE_SIZE];
-  int64_t* bufPtr = (int64_t*)writeBuf;
+  TqIdxPageBuf idxBuf;
+  int64_t* bufPtr = (int64_t*)idxBuf.buffer;
   TqMetaList *pHead = pMeta->unpersistHead;
   TqMetaList *pNode = pHead->unpersistNext;
   TqSerializedHead *pSHead = malloc(sizeof(TqSerializedHead));
@@ -278,6 +311,17 @@ int32_t tqStorePersist(TqMetaStore* pMeta) {
   pSHead->ssize = sizeof(TqSerializedHead);
   int allocatedSize = sizeof(TqSerializedHead);
   int offset = lseek(pMeta->fileFd, 0, SEEK_CUR);
+
+  tqReadLastPage(pMeta->idxFd, &idxBuf);
+
+  if(idxBuf.head.writeOffset == TQ_PAGE_SIZE) {
+    lseek(pMeta->idxFd, 0, SEEK_END);
+    memset(&idxBuf, 0, TQ_PAGE_SIZE);
+    idxBuf.head.writeOffset = TQ_IDX_PAGE_HEAD_SIZE;
+  } else {
+    bufPtr = POINTER_SHIFT(&idxBuf, idxBuf.head.writeOffset);
+  }
+
   while(pHead != pNode) {
     int nBytes = 0;
 
@@ -308,18 +352,23 @@ int32_t tqStorePersist(TqMetaStore* pMeta) {
       ASSERT(nBytesTxn == pSHead->ssize);
       nBytes += nBytesTxn;
     }
+    pNode->handle.offset = offset;
+    offset += nBytes;
 
     //write idx file
     //TODO: endian check and convert
     *(bufPtr++) = pNode->handle.key;
     *(bufPtr++) = pNode->handle.offset;
     *(bufPtr++) = (int64_t)nBytes;
-    if((char*)(bufPtr + 3) > writeBuf + TQ_PAGE_SIZE) {
-      nBytes = write(pMeta->idxFd, writeBuf, sizeof(writeBuf));
+    idxBuf.head.writeOffset += TQ_IDX_SIZE;
+
+    if(idxBuf.head.writeOffset >= TQ_PAGE_SIZE) {
+      nBytes = write(pMeta->idxFd, &idxBuf, TQ_PAGE_SIZE);
       //TODO: handle error with tfile
-      ASSERT(nBytes == sizeof(writeBuf));
-      memset(writeBuf, 0, TQ_PAGE_SIZE);
-      bufPtr = (int64_t*)writeBuf;
+      ASSERT(nBytes == TQ_PAGE_SIZE);
+      memset(&idxBuf, 0, TQ_PAGE_SIZE);
+      idxBuf.head.writeOffset = TQ_IDX_PAGE_HEAD_SIZE;
+      bufPtr = (int64_t*)&idxBuf.buffer;
     }
     //remove from unpersist list
     pHead->unpersistNext = pNode->unpersistNext;
@@ -331,7 +380,7 @@ int32_t tqStorePersist(TqMetaStore* pMeta) {
     if(pNode->handle.valueInUse == TQ_DELETE_TOKEN &&
         pNode->handle.valueInTxn == NULL
         ) {
-      int bucketKey = pNode->handle.key & TQ_BUCKET_SIZE;
+      int bucketKey = pNode->handle.key & TQ_BUCKET_MASK;
       TqMetaList* pBucketHead = pMeta->bucket[bucketKey];
       if(pBucketHead == pNode) {
         pMeta->bucket[bucketKey] = pNode->next;
@@ -351,11 +400,11 @@ int32_t tqStorePersist(TqMetaStore* pMeta) {
 
   //write left bytes
   free(pSHead);
-  if((char*)bufPtr != writeBuf) {
-    int used = (char*)bufPtr - writeBuf;
-    int nBytes = write(pMeta->idxFd, writeBuf, used);
+  //TODO: write new version in tfile
+  if((char*)bufPtr != idxBuf.buffer) {
+    int nBytes = write(pMeta->idxFd, &idxBuf, idxBuf.head.writeOffset);
     //TODO: handle error in tfile
-    ASSERT(nBytes == used);
+    ASSERT(nBytes == idxBuf.head.writeOffset);
   }
   //TODO: using fsync in tfile
   fsync(pMeta->idxFd);
@@ -364,7 +413,7 @@ int32_t tqStorePersist(TqMetaStore* pMeta) {
 }
 
 static int32_t tqHandlePutCommitted(TqMetaStore* pMeta, int64_t key, void* value) {
-  int64_t bucketKey = key & TQ_BUCKET_SIZE;
+  int64_t bucketKey = key & TQ_BUCKET_MASK;
   TqMetaList* pNode = pMeta->bucket[bucketKey];
   while(pNode) {
     if(pNode->handle.key == key) {
@@ -397,11 +446,12 @@ static int32_t tqHandlePutCommitted(TqMetaStore* pMeta, int64_t key, void* value
 }
 
 void* tqHandleGet(TqMetaStore* pMeta, int64_t key) {
-  int64_t bucketKey = key & TQ_BUCKET_SIZE;
+  int64_t bucketKey = key & TQ_BUCKET_MASK;
   TqMetaList* pNode = pMeta->bucket[bucketKey];
   while(pNode) {
     if(pNode->handle.key == key) {
-      if(pNode->handle.valueInUse != NULL) {
+      if(pNode->handle.valueInUse != NULL
+          && pNode->handle.valueInUse != TQ_DELETE_TOKEN) {
         return pNode->handle.valueInUse;
       } else {
         return NULL;
@@ -414,7 +464,7 @@ void* tqHandleGet(TqMetaStore* pMeta, int64_t key) {
 }
 
 int32_t tqHandleMovePut(TqMetaStore* pMeta, int64_t key, void* value) {
-  int64_t bucketKey = key & TQ_BUCKET_SIZE;
+  int64_t bucketKey = key & TQ_BUCKET_MASK;
   TqMetaList* pNode = pMeta->bucket[bucketKey];
   while(pNode) {
     if(pNode->handle.key == key) {
@@ -452,7 +502,7 @@ int32_t tqHandleCopyPut(TqMetaStore* pMeta, int64_t key, void* value, size_t vsi
     return -1;
   }
   memcpy(vmem, value, vsize);
-  int64_t bucketKey = key & TQ_BUCKET_SIZE;
+  int64_t bucketKey = key & TQ_BUCKET_MASK;
   TqMetaList* pNode = pMeta->bucket[bucketKey];
   while(pNode) {
     if(pNode->handle.key == key) {
@@ -484,7 +534,7 @@ int32_t tqHandleCopyPut(TqMetaStore* pMeta, int64_t key, void* value, size_t vsi
 }
 
 static void* tqHandleGetUncommitted(TqMetaStore* pMeta, int64_t key) {
-  int64_t bucketKey = key & TQ_BUCKET_SIZE;
+  int64_t bucketKey = key & TQ_BUCKET_MASK;
   TqMetaList* pNode = pMeta->bucket[bucketKey];
   while(pNode) {
     if(pNode->handle.key == key) {
@@ -502,10 +552,13 @@ static void* tqHandleGetUncommitted(TqMetaStore* pMeta, int64_t key) {
 }
 
 int32_t tqHandleCommit(TqMetaStore* pMeta, int64_t key) {
-  int64_t bucketKey = key & TQ_BUCKET_SIZE;
+  int64_t bucketKey = key & TQ_BUCKET_MASK;
   TqMetaList* pNode = pMeta->bucket[bucketKey];
   while(pNode) {
     if(pNode->handle.key == key) {
+      if(pNode->handle.valueInTxn == NULL) {
+        return -1;
+      }
       if(pNode->handle.valueInUse
           && pNode->handle.valueInUse != TQ_DELETE_TOKEN) {
         pMeta->deleter(pNode->handle.valueInUse);
@@ -518,11 +571,11 @@ int32_t tqHandleCommit(TqMetaStore* pMeta, int64_t key) {
       pNode = pNode->next;
     }
   }
-  return -1;
+  return -2;
 }
 
 int32_t tqHandleAbort(TqMetaStore* pMeta, int64_t key) {
-  int64_t bucketKey = key & TQ_BUCKET_SIZE;
+  int64_t bucketKey = key & TQ_BUCKET_MASK;
   TqMetaList* pNode = pMeta->bucket[bucketKey];
   while(pNode) {
     if(pNode->handle.key == key) {
@@ -543,12 +596,13 @@ int32_t tqHandleAbort(TqMetaStore* pMeta, int64_t key) {
 }
 
 int32_t tqHandleDel(TqMetaStore* pMeta, int64_t key) {
-  int64_t bucketKey = key & TQ_BUCKET_SIZE;
+  int64_t bucketKey = key & TQ_BUCKET_MASK;
   TqMetaList* pNode = pMeta->bucket[bucketKey];
   while(pNode) {
-      if(pNode->handle.valueInTxn
-          && pNode->handle.valueInTxn != TQ_DELETE_TOKEN) {
-      pMeta->deleter(pNode->handle.valueInTxn);
+    if(pNode->handle.valueInTxn != TQ_DELETE_TOKEN) {
+      if(pNode->handle.valueInTxn) {
+        pMeta->deleter(pNode->handle.valueInTxn);
+      }
       pNode->handle.valueInTxn = TQ_DELETE_TOKEN;
       tqLinkUnpersist(pMeta, pNode);
       return 0;
@@ -560,34 +614,7 @@ int32_t tqHandleDel(TqMetaStore* pMeta, int64_t key) {
   return -1;
 }
 
-int32_t tqHandleClear(TqMetaStore* pMeta, int64_t key) {
-  int64_t bucketKey = key & TQ_BUCKET_SIZE;
-  TqMetaList* pNode = pMeta->bucket[bucketKey];
-  bool exist = false;
-  while(pNode) {
-    if(pNode->handle.key == key) {
-      if(pNode->handle.valueInUse != NULL) {
-        exist = true;
-        if(pNode->handle.valueInUse != TQ_DELETE_TOKEN) {
-          pMeta->deleter(pNode->handle.valueInUse);
-        }
-        pNode->handle.valueInUse = TQ_DELETE_TOKEN;
-      }
-      if(pNode->handle.valueInTxn != NULL) {
-        exist = true;
-        if(pNode->handle.valueInTxn != TQ_DELETE_TOKEN) {
-          pMeta->deleter(pNode->handle.valueInTxn);
-        }
-        pNode->handle.valueInTxn = TQ_DELETE_TOKEN;
-      }
-      if(exist) {
-        tqLinkUnpersist(pMeta, pNode);
-        return 0;
-      }
-      return -1;
-    } else {
-      pNode = pNode->next;
-    }
-  }
-  return -2;
+//TODO: clean deleted idx and data from persistent file
+int32_t tqStoreCompact(TqMetaStore *pMeta) {
+  return 0;
 }

source/dnode/vnode/tq/test/tqMetaTest.cpp

@@ -57,6 +57,10 @@ TEST_F(TqMetaTest, copyPutTest) {
 
   Foo* pFoo = (Foo*) tqHandleGet(pMeta, 1);
   EXPECT_EQ(pFoo == NULL, true);
+
+  tqHandleCommit(pMeta, 1);
+  pFoo = (Foo*) tqHandleGet(pMeta, 1);
+  EXPECT_EQ(pFoo->a, 3);
 }
 
 TEST_F(TqMetaTest, persistTest) {
@@ -82,8 +86,6 @@ TEST_F(TqMetaTest, persistTest) {
 
   pBar = (Foo*)tqHandleGet(pMeta, 2);
   EXPECT_EQ(pBar == NULL, true);
-
-  //taosRemoveDir(pathName);
 }
 
 TEST_F(TqMetaTest, uncommittedTest) {
@@ -130,4 +132,163 @@ TEST_F(TqMetaTest, deleteTest) {
   tqHandleCommit(pMeta, 1);
   pFoo = (Foo*) tqHandleGet(pMeta, 1);
   EXPECT_EQ(pFoo == NULL, true);
+
+  tqStoreClose(pMeta);
+  pMeta = tqStoreOpen(pathName,
+      FooSerializer, FooDeserializer, FooDeleter);
+  ASSERT(pMeta);
+
+  pFoo = (Foo*) tqHandleGet(pMeta, 1);
+  EXPECT_EQ(pFoo == NULL, true);
+}
+
+TEST_F(TqMetaTest, intxnPersist) {
+  Foo* pFoo = (Foo*)malloc(sizeof(Foo));
+  pFoo->a = 3;
+  tqHandleMovePut(pMeta, 1, pFoo);
+  tqHandleCommit(pMeta, 1);
+
+  Foo* pBar = (Foo*)malloc(sizeof(Foo));
+  pBar->a = 4;
+  tqHandleMovePut(pMeta, 1, pBar);
+
+  Foo* pFoo1 = (Foo*)tqHandleGet(pMeta, 1);
+  EXPECT_EQ(pFoo1->a, 3);
+
+  tqStoreClose(pMeta);
+  pMeta = tqStoreOpen(pathName,
+      FooSerializer, FooDeserializer, FooDeleter);
+  ASSERT(pMeta);
+
+  pFoo1 = (Foo*)tqHandleGet(pMeta, 1);
+  EXPECT_EQ(pFoo1->a, 3);
+
+  tqHandleCommit(pMeta, 1);
+
+  pFoo1 = (Foo*)tqHandleGet(pMeta, 1);
+  EXPECT_EQ(pFoo1->a, 4);
+
+  tqStoreClose(pMeta);
+  pMeta = tqStoreOpen(pathName,
+      FooSerializer, FooDeserializer, FooDeleter);
+  ASSERT(pMeta);
+
+  pFoo1 = (Foo*)tqHandleGet(pMeta, 1);
+  EXPECT_EQ(pFoo1->a, 4);
+}
+
+TEST_F(TqMetaTest, multiplePage) {
+  srand(0);
+  std::vector<int> v;
+  for(int i = 0; i < 1000; i++) {
+    v.push_back(rand());
+    Foo foo;
+    foo.a = v[i];
+    tqHandleCopyPut(pMeta, i, &foo, sizeof(Foo));
+  }
+  for(int i = 0; i < 500; i++) {
+    tqHandleCommit(pMeta, i);
+    Foo* pFoo = (Foo*)tqHandleGet(pMeta, i);
+    ASSERT_EQ(pFoo != NULL, true) << " at idx " << i << "\n";
+    EXPECT_EQ(pFoo->a, v[i]);
+  }
+
+  tqStoreClose(pMeta);
+  pMeta = tqStoreOpen(pathName,
+      FooSerializer, FooDeserializer, FooDeleter);
+  ASSERT(pMeta);
+  
+  for(int i = 500; i < 1000; i++) {
+    tqHandleCommit(pMeta, i);
+    Foo* pFoo = (Foo*)tqHandleGet(pMeta, i);
+    ASSERT_EQ(pFoo != NULL, true) << " at idx " << i << "\n";
+    EXPECT_EQ(pFoo->a, v[i]);
+  }
+
+  for(int i = 0; i < 1000; i++) {
+    Foo* pFoo = (Foo*)tqHandleGet(pMeta, i);
+    ASSERT_EQ(pFoo != NULL, true) << " at idx " << i << "\n";
+    EXPECT_EQ(pFoo->a, v[i]);
+  }
+
+}
+
+TEST_F(TqMetaTest, multipleRewrite) {
+  srand(0);
+  std::vector<int> v;
+  for(int i = 0; i < 1000; i++) {
+    v.push_back(rand());
+    Foo foo;
+    foo.a = v[i];
+    tqHandleCopyPut(pMeta, i, &foo, sizeof(Foo));
+  }
+
+  for(int i = 0; i < 500; i++) {
+    tqHandleCommit(pMeta, i);
+    v[i] = rand();
+    Foo foo;
+    foo.a = v[i];
+    tqHandleCopyPut(pMeta, i, &foo, sizeof(Foo));
+  }
+
+  for(int i = 500; i < 1000; i++) {
+    v[i] = rand();
+    Foo foo;
+    foo.a = v[i];
+    tqHandleCopyPut(pMeta, i, &foo, sizeof(Foo));
+  }
+
+  for(int i = 0; i < 1000; i++) {
+    tqHandleCommit(pMeta, i);
+  }
+
+  tqStoreClose(pMeta);
+  pMeta = tqStoreOpen(pathName,
+      FooSerializer, FooDeserializer, FooDeleter);
+  ASSERT(pMeta);
+  
+  for(int i = 500; i < 1000; i++) {
+    v[i] = rand();
+    Foo foo;
+    foo.a = v[i];
+    tqHandleCopyPut(pMeta, i, &foo, sizeof(Foo));
+    tqHandleCommit(pMeta, i);
+  }
+
+  for(int i = 0; i < 1000; i++) {
+    Foo* pFoo = (Foo*)tqHandleGet(pMeta, i);
+    ASSERT_EQ(pFoo != NULL, true) << " at idx " << i << "\n";
+    EXPECT_EQ(pFoo->a, v[i]);
+  }
+
+}
+
+TEST_F(TqMetaTest, dupCommit) {
+  srand(0);
+  std::vector<int> v;
+  for(int i = 0; i < 1000; i++) {
+    v.push_back(rand());
+    Foo foo;
+    foo.a = v[i];
+    tqHandleCopyPut(pMeta, i, &foo, sizeof(Foo));
+  }
+
+  for(int i = 0; i < 1000; i++) {
+    int ret = tqHandleCommit(pMeta, i);
+    EXPECT_EQ(ret, 0);
+    ret = tqHandleCommit(pMeta, i);
+    EXPECT_EQ(ret, -1);
+  }
+
+  for(int i = 0; i < 1000; i++) {
+    int ret = tqHandleCommit(pMeta, i);
+    EXPECT_EQ(ret, -1);
+  }
+
+  for(int i = 0; i < 1000; i++) {
+    Foo* pFoo = (Foo*)tqHandleGet(pMeta, i);
+    ASSERT_EQ(pFoo != NULL, true) << " at idx " << i << "\n";
+    EXPECT_EQ(pFoo->a, v[i]);
+  }
+
 }

source/dnode/vnode/tsdb/inc/tsdbIdx.h

@@ -13,8 +13,8 @@
  * along with this program. If not, see <http://www.gnu.org/licenses/>.
  */
 
-#ifndef _TD_RAFT_INT_H_
-#define _TD_RAFT_INT_H_
+#ifndef _TD_TSDB_IDX_H_
+#define _TD_TSDB_IDX_H_
 
 #ifdef __cplusplus
 extern "C" {
@@ -24,4 +24,4 @@ extern "C" {
 }
 #endif
 
-#endif /*_TD_RAFT_INT_H_*/
+#endif /*_TD_TSDB_IDX_H_*/

source/dnode/vnode/tsdb/inc/tsdbOptions.h

@@ -20,6 +20,11 @@
 extern "C" {
 #endif
 
+extern const STsdbOptions defautlTsdbOptions;
+
+int  tsdbValidateOptions(const STsdbOptions *);
+void tsdbOptionsCopy(STsdbOptions *pDest, const STsdbOptions *pSrc);
+
 #ifdef __cplusplus
 }
 #endif

source/dnode/vnode/tsdb/inc/tsdbSMA.h

@@ -0,0 +1,27 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef _TD_TSDB_SMA_H_
+#define _TD_TSDB_SMA_H_
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /*_TD_TSDB_SMA_H_*/

source/dnode/vnode/tsdb/src/tsdbIdx.c

@@ -0,0 +1,14 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */

source/dnode/vnode/tsdb/src/tsdbMain.c

@@ -13,4 +13,82 @@
  * along with this program. If not, see <http://www.gnu.org/licenses/>.
  */
 
-#include "tsdbDef.h"
+#include "tsdbDef.h"
+
+static STsdb *tsdbNew(const char *path, const STsdbOptions *pTsdbOptions);
+static void   tsdbFree(STsdb *pTsdb);
+static int    tsdbOpenImpl(STsdb *pTsdb);
+static void   tsdbCloseImpl(STsdb *pTsdb);
+
+STsdb *tsdbOpen(const char *path, const STsdbOptions *pTsdbOptions) {
+  STsdb *pTsdb = NULL;
+
+  // Set default TSDB Options
+  if (pTsdbOptions == NULL) {
+    pTsdbOptions = &defautlTsdbOptions;
+  }
+
+  // Validate the options
+  if (tsdbValidateOptions(pTsdbOptions) < 0) {
+    // TODO: handle error
+    return NULL;
+  }
+
+  // Create the handle
+  pTsdb = tsdbNew(path, pTsdbOptions);
+  if (pTsdb == NULL) {
+    // TODO: handle error
+    return NULL;
+  }
+
+  taosMkDir(path);
+
+  // Open the TSDB
+  if (tsdbOpenImpl(pTsdb) < 0) {
+    // TODO: handle error
+    return NULL;
+  }
+
+  return pTsdb;
+}
+
+void tsdbClose(STsdb *pTsdb) {
+  if (pTsdb) {
+    tsdbCloseImpl(pTsdb);
+    tsdbFree(pTsdb);
+  }
+}
+
+void tsdbRemove(const char *path) { taosRemoveDir(path); }
+
+/* ------------------------ STATIC METHODS ------------------------ */
+static STsdb *tsdbNew(const char *path, const STsdbOptions *pTsdbOptions) {
+  STsdb *pTsdb = NULL;
+
+  pTsdb = (STsdb *)calloc(1, sizeof(STsdb));
+  if (pTsdb == NULL) {
+    // TODO: handle error
+    return NULL;
+  }
+
+  pTsdb->path = strdup(path);
+  tsdbOptionsCopy(&(pTsdb->options), pTsdbOptions);
+
+  return pTsdb;
+}
+
+static void tsdbFree(STsdb *pTsdb) {
+  if (pTsdb) {
+    tfree(pTsdb->path);
+    free(pTsdb);
+  }
+}
+
+static int tsdbOpenImpl(STsdb *pTsdb) {
+  // TODO
+  return 0;
+}
+
+static void tsdbCloseImpl(STsdb *pTsdb) {
+  // TODO
+}

source/dnode/vnode/tsdb/src/tsdbOptions.c

@@ -11,4 +11,24 @@
  *
  * You should have received a copy of the GNU Affero General Public License
  * along with this program. If not, see <http://www.gnu.org/licenses/>.
- */
+ */
+
+#include "tsdbDef.h"
+
+const STsdbOptions defautlTsdbOptions = {.lruCacheSize = 0};
+
+int tsdbOptionsInit(STsdbOptions *pTsdbOptions) {
+  // TODO
+  return 0;
+}
+
+void tsdbOptionsClear(STsdbOptions *pTsdbOptions) {
+  // TODO
+}
+
+int tsdbValidateOptions(const STsdbOptions *pTsdbOptions) {
+  // TODO
+  return 0;
+}
+
+void tsdbOptionsCopy(STsdbOptions *pDest, const STsdbOptions *pSrc) { memcpy(pDest, pSrc, sizeof(STsdbOptions)); }

source/dnode/vnode/tsdb/src/tsdbSMA.c

@@ -0,0 +1,14 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */

source/libs/sync/CMakeLists.txt

@@ -4,6 +4,7 @@ add_library(sync ${SYNC_SRC})
 target_link_libraries(
     sync
     PUBLIC common
+    PUBLIC transport
     PUBLIC util
     PUBLIC wal
 )

source/libs/sync/inc/raft.h

@@ -0,0 +1,144 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <cli@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef _TD_LIBS_SYNC_RAFT_H
+#define _TD_LIBS_SYNC_RAFT_H
+
+#include "sync.h"
+#include "sync_type.h"
+#include "raft_message.h"
+#include "sync_raft_impl.h"
+#include "sync_raft_quorum.h"
+
+typedef struct RaftLeaderState {
+
+} RaftLeaderState;
+
+typedef struct RaftCandidateState {
+  /* true if in pre-vote phase */
+  bool inPreVote;
+} RaftCandidateState;
+
+typedef struct SSyncRaftIOMethods {
+  // send SSyncMessage to node
+  int (*send)(const SSyncMessage* pMsg, const SNodeInfo* pNode);
+} SSyncRaftIOMethods;
+
+typedef int   (*SyncRaftStepFp)(SSyncRaft* pRaft, const SSyncMessage* pMsg);
+typedef void  (*SyncRaftTickFp)(SSyncRaft* pRaft);
+
+struct SSyncRaft {
+  // owner sync node
+  SSyncNode* pNode;
+
+  SSyncCluster cluster; 
+
+  int selfIndex;
+  SyncNodeId selfId;
+  SyncGroupId selfGroupId;
+
+  SSyncRaftIOMethods io;
+
+  SSyncFSM      fsm;
+  SSyncLogStore logStore;
+  SStateManager stateManager;
+
+  union {
+    RaftLeaderState leaderState;
+    RaftCandidateState candidateState;
+  };
+
+  SyncTerm term;
+  SyncNodeId voteFor;
+
+  SSyncRaftLog *log;
+
+  int maxMsgSize;
+  SSyncRaftProgressTracker *tracker;
+
+  ESyncState state;
+
+  // isLearner is true if the local raft node is a learner.
+  bool isLearner;
+
+  /**
+   * the leader id
+  **/
+  SyncNodeId leaderId;
+
+	/**
+   * leadTransferee is id of the leader transfer target when its value is not zero.
+	 * Follow the procedure defined in raft thesis 3.10.
+   **/
+  SyncNodeId leadTransferee;
+
+	/**
+   * Only one conf change may be pending (in the log, but not yet
+	 * applied) at a time. This is enforced via pendingConfIndex, which
+	 * is set to a value >= the log index of the latest pending
+	 * configuration change (if any). Config changes are only allowed to
+	 * be proposed if the leader's applied index is greater than this
+	 * value.
+   **/
+  SyncIndex pendingConfigIndex;
+
+	/** 
+   * an estimate of the size of the uncommitted tail of the Raft log. Used to
+	 * prevent unbounded log growth. Only maintained by the leader. Reset on
+	 * term changes.
+   **/
+  uint32_t uncommittedSize;
+ 
+	/** 
+   * number of ticks since it reached last electionTimeout when it is leader
+	 * or candidate.
+	 * number of ticks since it reached last electionTimeout or received a
+	 * valid message from current leader when it is a follower.
+  **/
+  uint16_t electionElapsed;
+
+	/**
+   *  number of ticks since it reached last heartbeatTimeout.
+	 * only leader keeps heartbeatElapsed.
+   **/
+  uint16_t heartbeatElapsed;
+
+  bool preVote;
+  bool checkQuorum;
+
+  int heartbeatTimeout;
+  int electionTimeout;
+
+	/**
+   * randomizedElectionTimeout is a random number between
+	 * [electiontimeout, 2 * electiontimeout - 1]. It gets reset
+	 * when raft changes its state to follower or candidate.
+   **/
+	int randomizedElectionTimeout;
+	bool disableProposalForwarding;
+
+  // current tick count since start up
+  uint32_t currentTick;
+
+  SyncRaftStepFp stepFp;
+
+  SyncRaftTickFp tickFp;
+};
+
+int32_t syncRaftStart(SSyncRaft* pRaft, const SSyncInfo* pInfo);
+int32_t syncRaftStep(SSyncRaft* pRaft, const SSyncMessage* pMsg);
+int32_t syncRaftTick(SSyncRaft* pRaft);
+
+#endif /* _TD_LIBS_SYNC_RAFT_H */

source/libs/sync/inc/raft_configuration.h

@@ -0,0 +1,27 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <cli@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef _TD_LIBS_SYNC_RAFT_CONFIGURATION_H
+#define _TD_LIBS_SYNC_RAFT_CONFIGURATION_H
+
+#include "sync.h"
+#include "sync_type.h"
+
+// return -1 if cannot find this id
+int syncRaftConfigurationIndexOfNode(SSyncRaft *pRaft, SyncNodeId id);
+
+int syncRaftConfigurationVoterCount(SSyncRaft *pRaft);
+
+#endif  /* _TD_LIBS_SYNC_RAFT_CONFIGURATION_H */

source/libs/sync/inc/raft_log.h

@@ -0,0 +1,76 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <cli@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef _TD_LIBS_SYNC_RAFT_LOG_H
+#define _TD_LIBS_SYNC_RAFT_LOG_H
+
+#include "sync.h"
+#include "sync_type.h"
+
+typedef enum SyncEntryType {
+  SYNC_ENTRY_TYPE_LOG = 1,
+}SyncEntryType;
+
+struct SSyncRaftEntry {
+  SyncTerm term;
+
+  SyncIndex index;
+
+  SyncEntryType type;
+
+  SSyncBuffer buffer;
+};
+
+struct SSyncRaftLog {
+  SyncIndex uncommittedConfigIndex;
+
+  SyncIndex commitIndex;
+
+  SyncIndex appliedIndex;
+
+
+};
+
+SSyncRaftLog* syncRaftLogOpen();
+
+SyncIndex syncRaftLogLastIndex(SSyncRaftLog* pLog);
+
+SyncIndex syncRaftLogSnapshotIndex(SSyncRaftLog* pLog);
+
+SyncTerm syncRaftLogLastTerm(SSyncRaftLog* pLog);
+
+bool syncRaftLogIsUptodate(SSyncRaftLog* pLog, SyncIndex index, SyncTerm term);
+
+int syncRaftLogNumOfPendingConf(SSyncRaftLog* pLog);
+
+bool syncRaftHasUnappliedLog(SSyncRaftLog* pLog);
+
+SyncTerm syncRaftLogTermOf(SSyncRaftLog* pLog, SyncIndex index);
+
+int syncRaftLogAppend(SSyncRaftLog* pLog, SSyncRaftEntry *pEntries, int n);
+
+int syncRaftLogAcquire(SSyncRaftLog* pLog, SyncIndex index, int maxMsgSize,
+                      SSyncRaftEntry **ppEntries, int *n);
+
+void syncRaftLogRelease(SSyncRaftLog* pLog, SyncIndex index,
+                      SSyncRaftEntry *pEntries, int n);
+
+bool syncRaftLogMatchTerm();
+
+static FORCE_INLINE bool syncRaftLogIsCommitted(SSyncRaftLog* pLog, SyncIndex index) {
+  return pLog->commitIndex > index;
+}
+
+#endif  /* _TD_LIBS_SYNC_RAFT_LOG_H */

source/libs/sync/inc/raft_message.h

@@ -0,0 +1,237 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <cli@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef _TD_LIBS_SYNC_RAFT_MESSAGE_H
+#define _TD_LIBS_SYNC_RAFT_MESSAGE_H
+
+#include "sync.h"
+#include "sync_type.h"
+
+/** 
+ * below define message type which handled by Raft.
+ * 
+ * internal message, which communicate between threads, start with RAFT_MSG_INTERNAL_*.
+ * internal message use pointer only and stack memory, need not to be decode/encode and free.
+ * 
+ * outter message start with RAFT_MSG_*, which communicate between cluster peers,
+ * need to implement its decode/encode functions.
+ **/
+typedef enum RaftMessageType {
+  // client propose a cmd
+  RAFT_MSG_INTERNAL_PROP = 1,
+
+  // node election timeout
+  RAFT_MSG_INTERNAL_ELECTION = 2,
+
+  RAFT_MSG_VOTE = 3,
+  RAFT_MSG_VOTE_RESP = 4,
+
+  RAFT_MSG_APPEND = 5,
+  RAFT_MSG_APPEND_RESP = 6,
+} RaftMessageType;
+
+typedef struct RaftMsgInternal_Prop {
+  const SSyncBuffer *pBuf;
+  bool isWeak;
+  void* pData;
+} RaftMsgInternal_Prop;
+
+typedef struct RaftMsgInternal_Election {
+
+} RaftMsgInternal_Election;
+
+typedef struct RaftMsg_Vote {
+  SyncRaftElectionType cType;
+  SyncIndex lastIndex;
+  SyncTerm lastTerm;
+} RaftMsg_Vote;
+
+typedef struct RaftMsg_VoteResp {
+  bool rejected;
+  SyncRaftElectionType cType;
+} RaftMsg_VoteResp;
+
+typedef struct RaftMsg_Append_Entries {
+  // index of log entry preceeding new ones
+  SyncIndex index;
+
+  // term of entry at prevIndex
+  SyncTerm term;
+
+  // leader's commit index.
+  SyncIndex commitIndex;
+
+  // size of the log entries array
+  int nEntries;
+
+  // log entries array
+  SSyncRaftEntry* entries;
+} RaftMsg_Append_Entries;
+
+typedef struct RaftMsg_Append_Resp {
+  SyncIndex index;
+} RaftMsg_Append_Resp;
+
+typedef struct SSyncMessage {
+  RaftMessageType msgType;
+  SyncTerm term;
+  SyncGroupId groupId;
+  SyncNodeId from;
+
+  union {
+    RaftMsgInternal_Prop propose;
+
+    RaftMsgInternal_Election election;
+
+    RaftMsg_Vote vote;
+    RaftMsg_VoteResp voteResp;
+
+    RaftMsg_Append_Entries appendEntries;
+    RaftMsg_Append_Resp appendResp;
+  };
+} SSyncMessage;
+
+static FORCE_INLINE SSyncMessage* syncInitPropMsg(SSyncMessage* pMsg, const SSyncBuffer* pBuf, void* pData, bool isWeak) {
+  *pMsg = (SSyncMessage) {
+    .msgType = RAFT_MSG_INTERNAL_PROP,
+    .term = 0,
+    .propose = (RaftMsgInternal_Prop) {
+      .isWeak = isWeak,
+      .pBuf = pBuf,
+      .pData = pData,
+    },
+  };
+
+  return pMsg;
+}
+
+static FORCE_INLINE SSyncMessage* syncInitElectionMsg(SSyncMessage* pMsg, SyncNodeId from) {
+  *pMsg = (SSyncMessage) {
+    .msgType = RAFT_MSG_INTERNAL_ELECTION,
+    .term = 0,
+    .from = from,
+    .election = (RaftMsgInternal_Election) {
+
+    },
+  };
+
+  return pMsg;
+}
+
+static FORCE_INLINE SSyncMessage* syncNewVoteMsg(SyncGroupId groupId, SyncNodeId from,
+                                                SyncTerm term, SyncRaftElectionType cType, 
+                                                SyncIndex lastIndex, SyncTerm lastTerm) {
+  SSyncMessage* pMsg = (SSyncMessage*)malloc(sizeof(SSyncMessage));
+  if (pMsg == NULL) {
+    return NULL;
+  }
+  *pMsg = (SSyncMessage) {
+    .groupId = groupId,
+    .from = from,
+    .term = term,
+    .msgType = RAFT_MSG_VOTE,
+    .vote = (RaftMsg_Vote) {
+      .cType = cType,
+      .lastIndex = lastIndex,
+      .lastTerm = lastTerm,
+    },
+  };
+
+  return pMsg;
+}
+
+static FORCE_INLINE SSyncMessage* syncNewVoteRespMsg(SyncGroupId groupId, SyncNodeId from,
+                                                  SyncRaftElectionType cType, bool rejected) {
+  SSyncMessage* pMsg = (SSyncMessage*)malloc(sizeof(SSyncMessage));
+  if (pMsg == NULL) {
+    return NULL;
+  }
+  *pMsg = (SSyncMessage) {
+    .groupId = groupId,
+    .from = from,
+    .msgType = RAFT_MSG_VOTE_RESP,
+    .voteResp = (RaftMsg_VoteResp) {
+      .cType = cType,
+      .rejected = rejected,
+    },
+  };
+
+  return pMsg;
+}
+
+static FORCE_INLINE SSyncMessage* syncNewAppendMsg(SyncGroupId groupId, SyncNodeId from,
+                                                  SyncTerm term, SyncIndex logIndex, SyncTerm logTerm,
+                                                  SyncIndex commitIndex, int nEntries, SSyncRaftEntry* entries) {
+  SSyncMessage* pMsg = (SSyncMessage*)malloc(sizeof(SSyncMessage));
+  if (pMsg == NULL) {
+    return NULL;
+  }
+  *pMsg = (SSyncMessage) {
+    .groupId = groupId,
+    .from = from,
+    .term = term,
+    .msgType = RAFT_MSG_APPEND,
+    .appendEntries = (RaftMsg_Append_Entries) {
+      .index = logIndex,
+      .term = logTerm,
+      .commitIndex = commitIndex,
+      .nEntries = nEntries,
+      .entries = entries,
+    },
+  };
+
+  return pMsg;
+}
+
+static FORCE_INLINE SSyncMessage* syncNewEmptyAppendRespMsg(SyncGroupId groupId, SyncNodeId from, SyncTerm term) {
+  SSyncMessage* pMsg = (SSyncMessage*)malloc(sizeof(SSyncMessage));
+  if (pMsg == NULL) {
+    return NULL;
+  }
+  *pMsg = (SSyncMessage) {
+    .groupId = groupId,
+    .from = from,
+    .term = term,
+    .msgType = RAFT_MSG_APPEND_RESP,
+    .appendResp = (RaftMsg_Append_Resp) {
+
+    },
+  };
+
+  return pMsg;
+}
+
+static FORCE_INLINE bool syncIsInternalMsg(RaftMessageType msgType) {
+  return msgType == RAFT_MSG_INTERNAL_PROP ||
+         msgType == RAFT_MSG_INTERNAL_ELECTION;
+}
+
+static FORCE_INLINE bool syncIsPreVoteRespMsg(const SSyncMessage* pMsg) {
+  return pMsg->msgType == RAFT_MSG_VOTE_RESP && pMsg->voteResp.cType == SYNC_RAFT_CAMPAIGN_PRE_ELECTION;
+}
+
+static FORCE_INLINE bool syncIsPreVoteMsg(const SSyncMessage* pMsg) {
+  return pMsg->msgType == RAFT_MSG_VOTE && pMsg->voteResp.cType == SYNC_RAFT_CAMPAIGN_PRE_ELECTION;
+}
+
+void syncFreeMessage(const SSyncMessage* pMsg);
+
+// message handlers
+int syncRaftHandleElectionMessage(SSyncRaft* pRaft, const SSyncMessage* pMsg);
+int syncRaftHandleVoteMessage(SSyncRaft* pRaft, const SSyncMessage* pMsg);
+int syncRaftHandleVoteRespMessage(SSyncRaft* pRaft, const SSyncMessage* pMsg);
+int syncRaftHandleAppendEntriesMessage(SSyncRaft* pRaft, const SSyncMessage* pMsg);
+
+#endif  /* _TD_LIBS_SYNC_RAFT_MESSAGE_H */

source/libs/sync/inc/raft_replication.h

@@ -0,0 +1,25 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <cli@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef TD_SYNC_RAFT_REPLICATION_H
+#define TD_SYNC_RAFT_REPLICATION_H
+
+#include "sync.h"
+#include "syncInt.h"
+#include "sync_type.h"
+
+int syncRaftReplicate(SSyncRaft* pRaft, int i);
+
+#endif  /* TD_SYNC_RAFT_REPLICATION_H */

source/libs/sync/inc/raft_unstable_log.h

@@ -0,0 +1,115 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <cli@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef TD_SYNC_RAFT_UNSTABLE_LOG_H
+#define TD_SYNC_RAFT_UNSTABLE_LOG_H
+
+#include "sync_type.h"
+
+/* in-memory unstable raft log storage */
+struct SSyncRaftUnstableLog {
+#if 0
+  /* Circular buffer of log entries */
+  RaftEntry *entries;
+
+  /* size of Circular buffer */
+  int size;
+
+  /* Indexes of used slots [front, back) */
+  int front, back;
+
+  /* Index of first entry is offset + 1 */
+  SyncIndex offset;
+
+  /* meta data of snapshot */
+  SSyncRaftUnstableLog snapshot;
+#endif
+};
+
+/**
+ * return index of last in memory log, return 0 if log is empty
+ **/
+//SyncIndex syncRaftLogLastIndex(SSyncRaftUnstableLog* pLog);
+
+#if 0
+void raftLogInit(RaftLog* pLog);
+
+void raftLogClose(RaftLog* pLog);
+
+/**
+ * When startup populating log entrues loaded from disk,
+ * init raft memory log with snapshot index,term and log start idnex.
+ **/ 
+/*
+void raftLogStart(RaftLog* pLog,
+                  RaftSnapshotMeta snapshot,
+                  SyncIndex startIndex);
+*/
+/** 
+ * Get the number of entries the log. 
+ **/
+int raftLogNumEntries(const RaftLog* pLog);
+
+
+
+/**
+ * return last term of in memory log, return 0 if log is empty
+ **/
+SyncTerm raftLogLastTerm(RaftLog* pLog);
+
+/**
+ * return term of log with the given index, return 0 if the term of index cannot be found
+ * , errCode will save the error code.
+ **/
+SyncTerm raftLogTermOf(RaftLog* pLog, SyncIndex index, RaftCode* errCode);
+
+/** 
+ * Get the last index of the most recent snapshot. Return 0 if there are no *
+ * snapshots. 
+ **/
+SyncIndex raftLogSnapshotIndex(RaftLog* pLog);
+
+/* Append a new entry to the log. */
+int raftLogAppend(RaftLog* pLog,
+                  SyncTerm term,
+                  const SSyncBuffer *buf);
+
+/**
+ * acquire log from given index onwards.
+ **/ 
+/*
+int raftLogAcquire(RaftLog* pLog,
+                  SyncIndex index,
+                  RaftEntry **ppEntries,
+                  int *n);
+
+void raftLogRelease(RaftLog* pLog,
+                    SyncIndex index,
+                    RaftEntry *pEntries,
+                    int n);
+*/
+/* Delete all entries from the given index (included) onwards. */
+void raftLogTruncate(RaftLog* pLog, SyncIndex index);
+
+/** 
+ * when taking a new snapshot, the function will update the last snapshot information and delete
+ * all entries up last_index - trailing (included). If the log contains no entry
+ * a last_index - trailing, then no entry will be deleted. 
+ **/
+void raftLogSnapshot(RaftLog* pLog, SyncIndex index, SyncIndex trailing);
+
+#endif
+
+#endif /* TD_SYNC_RAFT_UNSTABLE_LOG_H */

source/libs/sync/inc/syncInt.h

@@ -0,0 +1,71 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <cli@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef _TD_LIBS_SYNC_INT_H
+#define _TD_LIBS_SYNC_INT_H
+
+#include "thash.h"
+#include "os.h"
+#include "sync.h"
+#include "sync_type.h"
+#include "raft.h"
+#include "tlog.h"
+
+#define TAOS_SYNC_MAX_WORKER 3
+
+typedef struct SSyncWorker {
+  pthread_t thread;
+} SSyncWorker;
+
+struct SSyncNode {
+  pthread_mutex_t   mutex;
+  int32_t      refCount;
+  SyncGroupId   vgId;
+  SSyncRaft raft;
+  void* syncTimer;
+};
+
+typedef struct SSyncManager {
+  pthread_mutex_t   mutex;
+
+  // sync server rpc
+  void* serverRpc;
+  // rpc server hash table base on FQDN:port key
+  SHashObj* rpcServerTable;
+
+  // sync client rpc
+  void* clientRpc;
+
+  // worker threads
+  SSyncWorker worker[TAOS_SYNC_MAX_WORKER];
+
+  // vgroup hash table
+  SHashObj* vgroupTable;
+
+  // timer manager
+  void* syncTimerManager;
+
+} SSyncManager;
+
+extern SSyncManager* gSyncManager;
+
+#define syncFatal(...) do { if (sDebugFlag & DEBUG_FATAL) { taosPrintLog("SYNC FATAL ", 255, __VA_ARGS__); }}     while(0)
+#define syncError(...) do { if (sDebugFlag & DEBUG_ERROR) { taosPrintLog("SYNC ERROR ", 255, __VA_ARGS__); }}     while(0)
+#define syncWarn(...)  do { if (sDebugFlag & DEBUG_WARN)  { taosPrintLog("SYNC WARN ", 255, __VA_ARGS__); }}      while(0)
+#define syncInfo(...)  do { if (sDebugFlag & DEBUG_INFO)  { taosPrintLog("SYNC ", 255, __VA_ARGS__); }}           while(0)
+#define syncDebug(...) do { if (sDebugFlag & DEBUG_DEBUG) { taosPrintLog("SYNC ", sDebugFlag, __VA_ARGS__); }} while(0)
+#define syncTrace(...) do { if (sDebugFlag & DEBUG_TRACE) { taosPrintLog("SYNC ", sDebugFlag, __VA_ARGS__); }} while(0)
+
+#endif  /* _TD_LIBS_SYNC_INT_H */

source/libs/sync/inc/sync_raft_impl.h

@@ -0,0 +1,42 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <cli@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef _TD_LIBS_SYNC_RAFT_IMPL_H
+#define _TD_LIBS_SYNC_RAFT_IMPL_H
+
+#include "sync.h"
+#include "sync_type.h"
+#include "raft_message.h"
+#include "sync_raft_quorum.h"
+
+void syncRaftBecomeFollower(SSyncRaft* pRaft, SyncTerm term, SyncNodeId leaderId);
+void syncRaftBecomePreCandidate(SSyncRaft* pRaft);
+void syncRaftBecomeCandidate(SSyncRaft* pRaft);
+void syncRaftBecomeLeader(SSyncRaft* pRaft);
+
+void syncRaftStartElection(SSyncRaft* pRaft, SyncRaftElectionType cType);
+
+void syncRaftTriggerHeartbeat(SSyncRaft* pRaft);
+
+void syncRaftRandomizedElectionTimeout(SSyncRaft* pRaft);
+bool syncRaftIsPromotable(SSyncRaft* pRaft);
+bool syncRaftIsPastElectionTimeout(SSyncRaft* pRaft);
+int  syncRaftQuorum(SSyncRaft* pRaft);
+
+SSyncRaftVoteResult  syncRaftPollVote(SSyncRaft* pRaft, SyncNodeId id, 
+                                    bool preVote, bool accept, 
+                                    int* rejectNum, int *granted);
+
+#endif /* _TD_LIBS_SYNC_RAFT_IMPL_H */

source/libs/sync/inc/sync_raft_inflights.h

@@ -0,0 +1,77 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <cli@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http: *www.gnu.org/licenses/>.
+ */
+
+#ifndef TD_SYNC_RAFT_INFLIGHTS_H
+#define TD_SYNC_RAFT_INFLIGHTS_H
+
+#include "sync.h"
+
+/**
+ * SSyncRaftInflights limits the number of MsgApp (represented by the largest index
+ * contained within) sent to followers but not yet acknowledged by them. Callers
+ * use syncRaftInflightFull() to check whether more messages can be sent, 
+ * call syncRaftInflightAdd() whenever they are sending a new append, 
+ * and release "quota" via FreeLE() whenever an ack is received.
+**/
+typedef struct SSyncRaftInflights {
+  /* the starting index in the buffer */
+  int start;
+
+  /* number of inflights in the buffer */
+  int count;
+
+  /* the size of the buffer */
+  int size;
+
+	/** 
+   * buffer contains the index of the last entry
+	 * inside one message.
+   **/
+  SyncIndex* buffer;
+} SSyncRaftInflights;
+
+SSyncRaftInflights* syncRaftOpenInflights(int size);
+void syncRaftCloseInflights(SSyncRaftInflights*);
+
+static FORCE_INLINE void syncRaftInflightReset(SSyncRaftInflights* inflights) {  
+  inflights->count = 0;
+  inflights->start = 0;
+}
+
+static FORCE_INLINE bool syncRaftInflightFull(SSyncRaftInflights* inflights) {
+  return inflights->count == inflights->size;
+}
+
+/**
+ * syncRaftInflightAdd notifies the Inflights that a new message with the given index is being
+ * dispatched. syncRaftInflightFull() must be called prior to syncRaftInflightAdd() 
+ * to verify that there is room for one more message, 
+ * and consecutive calls to add syncRaftInflightAdd() must provide a
+ * monotonic sequence of indexes.
+ **/
+void syncRaftInflightAdd(SSyncRaftInflights* inflights, SyncIndex inflightIndex);
+
+/**
+ * syncRaftInflightFreeLE frees the inflights smaller or equal to the given `to` flight.
+ **/
+void syncRaftInflightFreeLE(SSyncRaftInflights* inflights, SyncIndex toIndex);
+
+/** 
+ * syncRaftInflightFreeFirstOne releases the first inflight. 
+ * This is a no-op if nothing is inflight.
+ **/
+void syncRaftInflightFreeFirstOne(SSyncRaftInflights* inflights);
+
+#endif /* TD_SYNC_RAFT_INFLIGHTS_H */

source/libs/sync/inc/sync_raft_progress.h

@@ -0,0 +1,237 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <cli@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http: *www.gnu.org/licenses/>.
+ */
+
+#ifndef TD_SYNC_RAFT_PROGRESS_H
+#define TD_SYNC_RAFT_PROGRESS_H
+
+#include "sync_type.h"
+#include "sync_raft_inflights.h"
+
+/** 
+ * State defines how the leader should interact with the follower.
+ *
+ * When in PROGRESS_STATE_PROBE, leader sends at most one replication message
+ * per heartbeat interval. It also probes actual progress of the follower.
+ * 
+ * When in PROGRESS_STATE_REPLICATE, leader optimistically increases next
+ * to the latest entry sent after sending replication message. This is
+ * an optimized state for fast replicating log entries to the follower.
+ *
+ * When in PROGRESS_STATE_SNAPSHOT, leader should have sent out snapshot
+ * before and stops sending any replication message.
+ * 
+ * PROGRESS_STATE_PROBE is the initial state.
+ **/
+typedef enum RaftProgressState {
+  /**
+	* StateProbe indicates a follower whose last index isn't known. Such a
+	* follower is "probed" (i.e. an append sent periodically) to narrow down
+	* its last index. In the ideal (and common) case, only one round of probing
+	* is necessary as the follower will react with a hint. Followers that are
+	* probed over extended periods of time are often offline.
+  **/
+  PROGRESS_STATE_PROBE = 0,
+
+  /**
+	* StateReplicate is the state steady in which a follower eagerly receives
+	* log entries to append to its log.
+  **/
+  PROGRESS_STATE_REPLICATE,
+
+  /**
+	* StateSnapshot indicates a follower that needs log entries not available
+	* from the leader's Raft log. Such a follower needs a full snapshot to
+	* return to StateReplicate.
+  **/
+  PROGRESS_STATE_SNAPSHOT,
+} RaftProgressState;
+
+/**
+ * Progress represents a follower’s progress in the view of the leader. Leader maintains
+ * progresses of all followers, and sends entries to the follower based on its progress.
+ **/
+struct SSyncRaftProgress {
+	SyncNodeId id;
+
+  SyncIndex nextIndex;
+
+  SyncIndex matchIndex;
+
+  /**
+	* State defines how the leader should interact with the follower.
+	*
+	* When in StateProbe, leader sends at most one replication message
+	* per heartbeat interval. It also probes actual progress of the follower.
+	*
+	* When in StateReplicate, leader optimistically increases next
+	* to the latest entry sent after sending replication message. This is
+	* an optimized state for fast replicating log entries to the follower.
+	*
+	* When in StateSnapshot, leader should have sent out snapshot
+	* before and stops sending any replication message.
+  **/
+  RaftProgressState state;
+
+	/** 
+   * pendingSnapshotIndex is used in PROGRESS_STATE_SNAPSHOT.
+	 * If there is a pending snapshot, the pendingSnapshotIndex will be set to the
+	 * index of the snapshot. If pendingSnapshotIndex is set, the replication process of
+	 * this Progress will be paused. raft will not resend snapshot until the pending one
+	 * is reported to be failed.
+   **/
+  SyncIndex pendingSnapshotIndex;
+
+	/** 
+   * recentActive is true if the progress is recently active. Receiving any messages
+   * from the corresponding follower indicates the progress is active.
+	 * RecentActive can be reset to false after an election timeout.
+   **/
+  bool recentActive;
+
+  /**
+	* probeSent is used while this follower is in StateProbe. When probeSent is
+	* true, raft should pause sending replication message to this peer until
+	* probeSent is reset. See ProbeAcked() and IsPaused().
+  **/
+	bool probeSent;
+
+  /**
+	 * inflights is a sliding window for the inflight messages.
+	 * Each inflight message contains one or more log entries.
+	 * The max number of entries per message is defined in raft config as MaxSizePerMsg.
+	 * Thus inflight effectively limits both the number of inflight messages
+	 * and the bandwidth each Progress can use.
+	 * When inflights is Full, no more message should be sent.
+	 * When a leader sends out a message, the index of the last
+	 * entry should be added to inflights. The index MUST be added
+	 * into inflights in order.
+	 * When a leader receives a reply, the previous inflights should
+	 * be freed by calling inflights.FreeLE with the index of the last
+	 * received entry.
+  **/
+  SSyncRaftInflights* inflights;
+
+  /**
+   *  IsLearner is true if this progress is tracked for a learner.
+   **/
+  bool isLearner;
+};
+
+void syncRaftInitProgress(int i, SSyncRaft* pRaft, SSyncRaftProgress* progress);
+
+/**
+ * syncRaftProgressBecomeProbe transitions into StateProbe. Next is reset to Match+1 or,
+ * optionally and if larger, the index of the pending snapshot.
+ **/
+void syncRaftProgressBecomeProbe(SSyncRaftProgress* progress);
+
+/**
+ * syncRaftProgressBecomeReplicate transitions into StateReplicate, resetting Next to Match+1.
+ **/ 
+void syncRaftProgressBecomeReplicate(SSyncRaftProgress* progress);
+
+/**
+ * syncRaftProgressMaybeUpdate is called when an MsgAppResp arrives from the follower, with the
+ * index acked by it. The method returns false if the given n index comes from
+ * an outdated message. Otherwise it updates the progress and returns true.
+ **/
+bool syncRaftProgressMaybeUpdate(SSyncRaftProgress* progress, SyncIndex lastIndex);
+
+/**
+ * syncRaftProgressOptimisticNextIndex signals that appends all the way up to and including index n
+ * are in-flight. As a result, Next is increased to n+1.
+ **/
+static FORCE_INLINE void syncRaftProgressOptimisticNextIndex(SSyncRaftProgress* progress, SyncIndex nextIndex) {
+  progress->nextIndex = nextIndex + 1;
+}
+
+/**
+ * syncRaftProgressMaybeDecrTo adjusts the Progress to the receipt of a MsgApp rejection. The
+ * arguments are the index of the append message rejected by the follower, and
+ * the hint that we want to decrease to.
+ *
+ * Rejections can happen spuriously as messages are sent out of order or
+ * duplicated. In such cases, the rejection pertains to an index that the
+ * Progress already knows were previously acknowledged, and false is returned
+ * without changing the Progress.
+ *
+ * If the rejection is genuine, Next is lowered sensibly, and the Progress is
+ * cleared for sending log entries.
+**/
+bool syncRaftProgressMaybeDecrTo(SSyncRaftProgress* progress,
+                                SyncIndex rejected, SyncIndex matchHint);
+
+/**
+ * syncRaftProgressIsPaused returns whether sending log entries to this node has been throttled.
+ * This is done when a node has rejected recent MsgApps, is currently waiting
+ * for a snapshot, or has reached the MaxInflightMsgs limit. In normal
+ * operation, this is false. A throttled node will be contacted less frequently
+ * until it has reached a state in which it's able to accept a steady stream of
+ * log entries again.
+ **/
+bool syncRaftProgressIsPaused(SSyncRaftProgress* progress);
+
+static FORCE_INLINE SyncIndex syncRaftProgressNextIndex(SSyncRaftProgress* progress) {
+  return progress->nextIndex;
+}
+
+static FORCE_INLINE RaftProgressState syncRaftProgressInReplicate(SSyncRaftProgress* progress) {
+  return progress->state == PROGRESS_STATE_REPLICATE;
+}
+
+static FORCE_INLINE RaftProgressState syncRaftProgressInSnapshot(SSyncRaftProgress* progress) {
+  return progress->state == PROGRESS_STATE_SNAPSHOT;
+}
+
+static FORCE_INLINE RaftProgressState syncRaftProgressInProbe(SSyncRaftProgress* progress) {
+  return progress->state == PROGRESS_STATE_PROBE;
+}
+
+static FORCE_INLINE bool syncRaftProgressRecentActive(SSyncRaftProgress* progress) {
+  return progress->recentActive;
+}
+
+/** 
+ * return true if progress's log is up-todate
+ **/
+bool syncRaftProgressIsUptodate(SSyncRaft* pRaft, SSyncRaftProgress* progress);
+
+void syncRaftProgressBecomeSnapshot(SSyncRaftProgress* progress, SyncIndex snapshotIndex);
+
+
+
+#if 0
+
+void syncRaftProgressAbortSnapshot(SSyncRaft* pRaft, int i);
+
+
+
+SyncIndex syncRaftProgressMatchIndex(SSyncRaft* pRaft, int i);
+
+void syncRaftProgressUpdateLastSend(SSyncRaft* pRaft, int i);
+
+void syncRaftProgressUpdateSnapshotLastSend(SSyncRaft* pRaft, int i);
+
+bool syncRaftProgressResetRecentRecv(SSyncRaft* pRaft, int i);
+
+void syncRaftProgressMarkRecentRecv(SSyncRaft* pRaft, int i);
+
+
+
+void syncRaftProgressAbortSnapshot(SSyncRaft* pRaft, int i);
+
+#endif
+
+#endif /* TD_SYNC_RAFT_PROGRESS_H */

source/libs/sync/inc/sync_raft_progress_tracker.h

@@ -0,0 +1,113 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <cli@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef _TD_LIBS_SYNC_RAFT_PROGRESS_TRACKER_H
+#define _TD_LIBS_SYNC_RAFT_PROGRESS_TRACKER_H
+
+#include "sync_type.h"
+#include "sync_raft_quorum_joint.h"
+#include "sync_raft_progress.h"
+
+struct SSyncRaftProgressTrackerConfig {
+  SSyncRaftQuorumJointConfig voters;
+
+	/** 
+	 * autoLeave is true if the configuration is joint and a transition to the
+	 * incoming configuration should be carried out automatically by Raft when
+	 * this is possible. If false, the configuration will be joint until the
+	 * application initiates the transition manually.
+   **/
+  bool autoLeave;
+
+	/** 
+   * Learners is a set of IDs corresponding to the learners active in the
+	 * current configuration.
+   * 
+	 * Invariant: Learners and Voters does not intersect, i.e. if a peer is in
+	 * either half of the joint config, it can't be a learner; if it is a
+	 * learner it can't be in either half of the joint config. This invariant
+	 * simplifies the implementation since it allows peers to have clarity about
+	 * its current role without taking into account joint consensus.
+   **/
+  SyncNodeId learners[TSDB_MAX_REPLICA];
+
+  /**
+   *  When we turn a voter into a learner during a joint consensus transition,
+	 * we cannot add the learner directly when entering the joint state. This is
+	 * because this would violate the invariant that the intersection of
+	 * voters and learners is empty. For example, assume a Voter is removed and
+	 * immediately re-added as a learner (or in other words, it is demoted):
+   * 
+	 * Initially, the configuration will be
+   * 
+	 *   voters:   {1 2 3}
+	 *   learners: {}
+   * 
+	 * and we want to demote 3. Entering the joint configuration, we naively get
+   * 
+	 *   voters:   {1 2} & {1 2 3}
+	 *   learners: {3}
+   * 
+	 * but this violates the invariant (3 is both voter and learner). Instead,
+	 * we get
+   * 
+	 *   voters:   {1 2} & {1 2 3}
+	 *   learners: {}
+	 *   next_learners: {3}
+   * 
+	 * Where 3 is now still purely a voter, but we are remembering the intention
+	 * to make it a learner upon transitioning into the final configuration:
+   * 
+	 *   voters:   {1 2}
+	 *   learners: {3}
+	 *   next_learners: {}
+   * 
+	 * Note that next_learners is not used while adding a learner that is not
+	 * also a voter in the joint config. In this case, the learner is added
+	 * right away when entering the joint configuration, so that it is caught up
+	 * as soon as possible.
+   **/
+  SyncNodeId learnersNext[TSDB_MAX_REPLICA];
+};
+
+struct SSyncRaftProgressTracker {
+  SSyncRaftProgressTrackerConfig config;
+
+  SSyncRaftProgress progressMap[TSDB_MAX_REPLICA];
+
+	SyncRaftVoteResult votes[TSDB_MAX_REPLICA];
+  int maxInflight;
+};
+
+SSyncRaftProgressTracker* syncRaftOpenProgressTracker();
+
+void syncRaftResetVotes(SSyncRaftProgressTracker*);
+
+typedef void (*visitProgressFp)(int i, SSyncRaftProgress* progress, void* arg);
+void syncRaftProgressVisit(SSyncRaftProgressTracker*, visitProgressFp visit, void* arg);
+
+/**
+ * syncRaftRecordVote records that the node with the given id voted for this Raft
+ * instance if v == true (and declined it otherwise).
+ **/
+void syncRaftRecordVote(SSyncRaftProgressTracker* tracker, int i, bool grant);
+
+/** 
+ * syncRaftTallyVotes returns the number of granted and rejected Votes, and whether the
+ * election outcome is known.
+ **/
+SyncRaftVoteResult syncRaftTallyVotes(SSyncRaftProgressTracker* tracker, int* rejected, int *granted);
+
+#endif  /* _TD_LIBS_SYNC_RAFT_PROGRESS_TRACKER_H */

source/libs/sync/inc/sync_raft_quorum.h

@@ -0,0 +1,40 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <cli@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef TD_SYNC_RAFT_QUORUM_H
+#define TD_SYNC_RAFT_QUORUM_H
+
+/**
+ * SSyncRaftVoteResult indicates the outcome of a vote.
+ **/
+typedef enum {
+	/**
+	 *  SYNC_RAFT_VOTE_PENDING indicates that the decision of the vote depends on future
+	 * votes, i.e. neither "yes" or "no" has reached quorum yet.
+	 **/
+	SYNC_RAFT_VOTE_PENDING = 1,
+
+	/** 
+	 * SYNC_RAFT_VOTE_LOST indicates that the quorum has voted "no".
+	 **/
+	SYNC_RAFT_VOTE_LOST = 2,
+
+	/** 
+	 * SYNC_RAFT_VOTE_WON indicates that the quorum has voted "yes".
+	 **/
+	SYNC_RAFT_VOTE_WON = 3,
+} SSyncRaftVoteResult;
+
+#endif /* TD_SYNC_RAFT_QUORUM_H */

source/libs/sync/inc/sync_raft_quorum_joint.h

@@ -0,0 +1,38 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <cli@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef _TD_LIBS_SYNC_RAFT_QUORUM_JOINT_H
+#define _TD_LIBS_SYNC_RAFT_QUORUM_JOINT_H
+
+#include "taosdef.h"
+#include "sync.h"
+#include "sync_type.h"
+
+/**
+ * SSyncRaftQuorumJointConfig is a configuration of two groups of (possibly overlapping)
+ * majority configurations. Decisions require the support of both majorities.
+ **/
+typedef struct SSyncRaftQuorumJointConfig {
+  SSyncCluster majorityConfig[2];
+}SSyncRaftQuorumJointConfig;
+
+/**
+ * syncRaftVoteResult takes a mapping of voters to yes/no (true/false) votes and returns
+ * a result indicating whether the vote is pending, lost, or won. A joint quorum
+ * requires both majority quorums to vote in favor.
+ **/
+SyncRaftVoteResult syncRaftVoteResult(SSyncRaftQuorumJointConfig* config, const SyncRaftVoteResult* votes);
+
+#endif /* _TD_LIBS_SYNC_RAFT_QUORUM_JOINT_H */

source/libs/sync/inc/sync_raft_quorum_majority.h

@@ -0,0 +1,30 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <cli@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef _TD_LIBS_SYNC_RAFT_QUORUM_MAJORITY_H
+#define _TD_LIBS_SYNC_RAFT_QUORUM_MAJORITY_H
+
+#include "sync.h"
+#include "sync_type.h"
+
+/**
+ * syncRaftMajorityVoteResult takes a mapping of voters to yes/no (true/false) votes and returns
+ * a result indicating whether the vote is pending (i.e. neither a quorum of
+ * yes/no has been reached), won (a quorum of yes has been reached), or lost (a
+ * quorum of no has been reached).
+ **/
+SyncRaftVoteResult syncRaftMajorityVoteResult(SSyncCluster* config, const SyncRaftVoteResult* votes);
+
+#endif /* _TD_LIBS_SYNC_RAFT_QUORUM_MAJORITY_H */

source/libs/sync/inc/sync_type.h

@@ -0,0 +1,66 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <cli@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef _TD_LIBS_SYNC_TYPE_H
+#define _TD_LIBS_SYNC_TYPE_H
+
+#include <stdint.h>
+#include "osMath.h"
+
+#define SYNC_NON_NODE_ID -1
+#define SYNC_NON_TERM     0
+
+typedef int32_t SyncTime;
+typedef uint32_t SyncTick;
+
+typedef struct SSyncRaft SSyncRaft;
+
+typedef struct SSyncRaftProgress SSyncRaftProgress;
+typedef struct SSyncRaftProgressTrackerConfig SSyncRaftProgressTrackerConfig;
+
+typedef struct SSyncRaftProgressTracker SSyncRaftProgressTracker;
+
+typedef struct SSyncRaftLog SSyncRaftLog;
+
+typedef struct SSyncRaftEntry SSyncRaftEntry;
+
+#if 0
+#ifndef MIN
+#define MIN(x, y) (((x) < (y)) ? (x) : (y))
+#endif
+
+#ifndef MAX
+#define MAX(x, y) (((x) > (y)) ? (x) : (y))
+#endif
+#endif
+
+typedef enum {
+  SYNC_RAFT_CAMPAIGN_PRE_ELECTION = 0,
+  SYNC_RAFT_CAMPAIGN_ELECTION     = 1,
+  SYNC_RAFT_CAMPAIGN_TRANSFER     = 2,
+} SyncRaftElectionType;
+
+typedef enum {
+  // the init vote resp status
+  SYNC_RAFT_VOTE_RESP_UNKNOWN = 0,
+
+  // grant the vote request
+  SYNC_RAFT_VOTE_RESP_GRANT   = 1,
+
+  //reject the vote request
+  SYNC_RAFT_VOTE_RESP_REJECT  = 2,
+} SyncRaftVoteResult;
+
+#endif  /* _TD_LIBS_SYNC_TYPE_H */

source/libs/sync/src/raft.c

@@ -0,0 +1,203 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <cli@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "raft.h"
+#include "raft_configuration.h"
+#include "raft_log.h"
+#include "raft_replication.h"
+#include "sync_raft_progress_tracker.h"
+#include "syncInt.h"
+
+#define RAFT_READ_LOG_MAX_NUM 100
+
+static bool preHandleMessage(SSyncRaft* pRaft, const SSyncMessage* pMsg);
+static bool preHandleNewTermMessage(SSyncRaft* pRaft, const SSyncMessage* pMsg);
+static bool preHandleOldTermMessage(SSyncRaft* pRaft, const SSyncMessage* pMsg);
+
+int32_t syncRaftStart(SSyncRaft* pRaft, const SSyncInfo* pInfo) {
+  SSyncNode* pNode = pRaft->pNode;
+  SSyncServerState serverState;
+  SStateManager* stateManager;
+  SSyncLogStore* logStore;
+  SSyncFSM* fsm;
+  SyncIndex initIndex = pInfo->snapshotIndex;
+  SSyncBuffer buffer[RAFT_READ_LOG_MAX_NUM];
+  int nBuf, limit, i;
+  
+  memset(pRaft, 0, sizeof(SSyncRaft));
+
+  memcpy(&pRaft->fsm, &pInfo->fsm, sizeof(SSyncFSM));
+  memcpy(&pRaft->logStore, &pInfo->logStore, sizeof(SSyncLogStore));
+  memcpy(&pRaft->stateManager, &pInfo->stateManager, sizeof(SStateManager));
+
+  stateManager = &(pRaft->stateManager);
+  logStore = &(pRaft->logStore);
+  fsm = &(pRaft->fsm);
+
+  // init progress tracker
+  pRaft->tracker = syncRaftOpenProgressTracker();
+  if (pRaft->tracker == NULL) {
+    return -1;
+  }
+
+  // open raft log
+  if ((pRaft->log = syncRaftLogOpen()) == NULL) {
+    return -1;
+  }
+  // read server state
+  if (stateManager->readServerState(stateManager, &serverState) != 0) {
+    syncError("readServerState for vgid %d fail", pInfo->vgId);
+    return -1;
+  }
+  assert(initIndex <= serverState.commitIndex);
+
+  // restore fsm state from snapshot index + 1 until commitIndex
+  ++initIndex;
+  while (initIndex <= serverState.commitIndex) {
+    limit = MIN(RAFT_READ_LOG_MAX_NUM, serverState.commitIndex - initIndex + 1);
+
+    if (logStore->logRead(logStore, initIndex, limit, buffer, &nBuf) != 0) {
+      return -1;
+    }
+    assert(limit == nBuf);
+  
+    for (i = 0; i < limit; ++i) {
+      fsm->applyLog(fsm, initIndex + i, &(buffer[i]), NULL);
+      free(buffer[i].data);
+    }
+    initIndex += nBuf;
+  }
+  assert(initIndex == serverState.commitIndex);
+
+  //pRaft->heartbeatTimeoutTick = 1;
+
+  syncRaftBecomeFollower(pRaft, pRaft->term, SYNC_NON_NODE_ID);
+
+  pRaft->selfIndex = pRaft->cluster.selfIndex;
+
+  syncInfo("[%d:%d] restore vgid %d state: snapshot index success", 
+    pRaft->selfGroupId, pRaft->selfId, pInfo->vgId);
+  return 0;
+}
+
+int32_t syncRaftStep(SSyncRaft* pRaft, const SSyncMessage* pMsg) {
+  syncDebug("from %d, type:%d, term:%" PRId64 ", state:%d",
+    pMsg->from, pMsg->msgType, pMsg->term, pRaft->state);
+
+  if (preHandleMessage(pRaft, pMsg)) {
+    syncFreeMessage(pMsg);
+    return 0;
+  }
+
+  RaftMessageType msgType = pMsg->msgType;
+  if (msgType == RAFT_MSG_INTERNAL_ELECTION) {
+    syncRaftHandleElectionMessage(pRaft, pMsg);
+  } else if (msgType == RAFT_MSG_VOTE) {
+    syncRaftHandleVoteMessage(pRaft, pMsg);
+  } else {
+    pRaft->stepFp(pRaft, pMsg);
+  }
+
+  syncFreeMessage(pMsg);
+  return 0;
+}
+
+int32_t syncRaftTick(SSyncRaft* pRaft) {
+  pRaft->currentTick += 1;
+  return 0;
+}
+
+/**
+ * pre-handle message, return true means no need to continue
+ * Handle the message term, which may result in our stepping down to a follower.
+ **/
+static bool preHandleMessage(SSyncRaft* pRaft, const SSyncMessage* pMsg) {
+  // local message?
+  if (pMsg->term == 0) {
+    return false;
+  }
+
+  if (pMsg->term > pRaft->term) {
+    return preHandleNewTermMessage(pRaft, pMsg);
+  } else if (pMsg->term < pRaft->term) {
+    return preHandleOldTermMessage(pRaft, pMsg);
+  }
+
+  return false;
+}
+
+static bool preHandleNewTermMessage(SSyncRaft* pRaft, const SSyncMessage* pMsg) {
+  SyncNodeId leaderId = pMsg->from;
+  RaftMessageType msgType = pMsg->msgType;
+
+  if (msgType == RAFT_MSG_VOTE) {
+    // TODO
+    leaderId = SYNC_NON_NODE_ID;
+  }
+
+  if (syncIsPreVoteMsg(pMsg)) {
+    // Never change our term in response to a PreVote
+  } else if (syncIsPreVoteRespMsg(pMsg) && !pMsg->voteResp.rejected) {
+		/**
+     * We send pre-vote requests with a term in our future. If the
+		 * pre-vote is granted, we will increment our term when we get a
+		 * quorum. If it is not, the term comes from the node that
+		 * rejected our vote so we should become a follower at the new
+		 * term.
+     **/
+  } else {
+    syncInfo("[%d:%d] [term:%" PRId64 "] received a %d message with higher term from %d [term:%" PRId64 "]",
+      pRaft->selfGroupId, pRaft->selfId, pRaft->term, msgType, pMsg->from, pMsg->term);
+    syncRaftBecomeFollower(pRaft, pMsg->term, leaderId);
+  }
+
+  return false;
+}
+
+static bool preHandleOldTermMessage(SSyncRaft* pRaft, const SSyncMessage* pMsg) {
+  if (pRaft->checkQuorum && pMsg->msgType == RAFT_MSG_APPEND) {
+		/**
+     * We have received messages from a leader at a lower term. It is possible
+		 * that these messages were simply delayed in the network, but this could
+		 * also mean that this node has advanced its term number during a network
+		 * partition, and it is now unable to either win an election or to rejoin
+	   * the majority on the old term. If checkQuorum is false, this will be
+		 * handled by incrementing term numbers in response to MsgVote with a
+		 * higher term, but if checkQuorum is true we may not advance the term on
+		 * MsgVote and must generate other messages to advance the term. The net
+		 * result of these two features is to minimize the disruption caused by
+		 * nodes that have been removed from the cluster's configuration: a
+		 * removed node will send MsgVotes (or MsgPreVotes) which will be ignored,
+		 * but it will not receive MsgApp or MsgHeartbeat, so it will not create
+		 * disruptive term increases
+    **/
+    int peerIndex = syncRaftConfigurationIndexOfNode(pRaft, pMsg->from);
+    if (peerIndex < 0) {
+      return true;
+    }
+    SSyncMessage* msg = syncNewEmptyAppendRespMsg(pRaft->selfGroupId, pRaft->selfId, pRaft->term);
+    if (msg == NULL) {
+      return true;
+    }
+
+    pRaft->io.send(msg, &(pRaft->cluster.nodeInfo[peerIndex]));
+  } else {
+    // ignore other cases
+    syncInfo("[%d:%d] [term:%" PRId64 "] ignored a %d message with lower term from %d [term:%" PRId64 "]",
+      pRaft->selfGroupId, pRaft->selfId, pRaft->term, pMsg->msgType, pMsg->from, pMsg->term);    
+  }
+
+  return true;
+}

source/libs/sync/src/raft_configuration.c

@@ -0,0 +1,25 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <cli@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "raft_configuration.h"
+#include "raft.h"
+
+int syncRaftConfigurationIndexOfNode(SSyncRaft *pRaft, SyncNodeId id) {
+  return (int)(id);
+}
+
+int syncRaftConfigurationVoterCount(SSyncRaft *pRaft) {
+  return pRaft->cluster.replica;
+}

source/libs/sync/src/raft_election.c

@@ -0,0 +1,81 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <cli@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "syncInt.h"
+#include "raft.h"
+#include "raft_log.h"
+#include "raft_message.h"
+
+void syncRaftStartElection(SSyncRaft* pRaft, SyncRaftElectionType cType) {
+  SyncTerm term;
+  bool preVote;
+  RaftMessageType voteMsgType;
+
+  if (syncRaftIsPromotable(pRaft)) {
+    syncDebug("[%d:%d] is unpromotable; campaign() should have been called", pRaft->selfGroupId, pRaft->selfId);
+    return 0;
+  }
+
+  if (cType == SYNC_RAFT_CAMPAIGN_PRE_ELECTION) {
+    syncRaftBecomePreCandidate(pRaft);
+    preVote = true;
+    // PreVote RPCs are sent for the next term before we've incremented r.Term.
+    term = pRaft->term + 1;
+  } else {
+    syncRaftBecomeCandidate(pRaft);
+    voteMsgType = RAFT_MSG_VOTE;
+    term = pRaft->term;
+    preVote = false;
+  }
+
+  int quorum = syncRaftQuorum(pRaft);
+  SSyncRaftVoteResult result = syncRaftPollVote(pRaft, pRaft->selfId, preVote, true, NULL, NULL);
+  if (result == SYNC_RAFT_VOTE_WON) {
+		/**
+     * We won the election after voting for ourselves (which must mean that
+		 * this is a single-node cluster). Advance to the next state.
+     **/
+    if (cType == SYNC_RAFT_CAMPAIGN_PRE_ELECTION) {
+      syncRaftStartElection(pRaft, SYNC_RAFT_CAMPAIGN_ELECTION);
+    } else {
+      syncRaftBecomeLeader(pRaft);
+    }
+    return;
+  }
+
+  // broadcast vote message to other peers
+  int i;
+  SyncIndex lastIndex = syncRaftLogLastIndex(pRaft->log);
+  SyncTerm lastTerm = syncRaftLogLastTerm(pRaft->log);
+  for (i = 0; i < pRaft->cluster.replica; ++i) {
+    if (i == pRaft->cluster.selfIndex) {
+      continue;
+    }
+
+    SyncNodeId nodeId = pRaft->cluster.nodeInfo[i].nodeId;
+
+    SSyncMessage* pMsg = syncNewVoteMsg(pRaft->selfGroupId, pRaft->selfId, 
+                                        term, cType, lastIndex, lastTerm);
+    if (pMsg == NULL) {
+      continue;
+    }
+
+    syncInfo("[%d:%d] [logterm: %" PRId64 ", index: %" PRId64 "] sent %d request to %d at term %" PRId64 "", 
+      pRaft->selfGroupId, pRaft->selfId, lastTerm, 
+      lastIndex, voteMsgType, nodeId, pRaft->term);
+
+    pRaft->io.send(pMsg, &(pRaft->cluster.nodeInfo[i]));
+  }
+}

source/libs/sync/src/raft_handle_append_entries_message.c

@@ -0,0 +1,49 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <cli@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "syncInt.h"
+#include "raft.h"
+#include "raft_log.h"
+#include "raft_configuration.h"
+#include "raft_message.h"
+
+int syncRaftHandleAppendEntriesMessage(SSyncRaft* pRaft, const SSyncMessage* pMsg) {
+  const RaftMsg_Append_Entries *appendEntries = &(pMsg->appendEntries);
+  
+  int peerIndex = syncRaftConfigurationIndexOfNode(pRaft, pMsg->from);
+  
+  if (peerIndex < 0) {
+    return 0;
+  }
+
+  SSyncMessage* pRespMsg = syncNewEmptyAppendRespMsg(pRaft->selfGroupId, pRaft->selfId, pRaft->term);
+  if (pRespMsg == NULL) {
+    return 0;
+  }
+
+  RaftMsg_Append_Entries *appendResp = &(pRespMsg->appendResp);
+  // ignore committed logs
+  if (syncRaftLogIsCommitted(pRaft->log, appendEntries->index)) {
+    appendResp->index = pRaft->log->commitIndex;
+    goto out;
+  }
+
+  syncInfo("[%d:%d] recv append from %d index %" PRId64"",
+      pRaft->selfGroupId, pRaft->selfId, pMsg->from, appendEntries->index); 
+
+out:
+  pRaft->io.send(pRespMsg, &(pRaft->cluster.nodeInfo[peerIndex]));
+  return 0;
+}

source/libs/sync/src/raft_handle_election_message.c

@@ -0,0 +1,47 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <cli@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "syncInt.h"
+#include "raft.h"
+#include "raft_log.h"
+#include "raft_message.h"
+
+int syncRaftHandleElectionMessage(SSyncRaft* pRaft, const SSyncMessage* pMsg) {
+  if (pRaft->state == TAOS_SYNC_STATE_LEADER) {
+    syncDebug("[%d:%d] ignoring RAFT_MSG_INTERNAL_ELECTION because already leader", pRaft->selfGroupId, pRaft->selfId);
+    return 0;
+  }
+
+  if (!syncRaftIsPromotable(pRaft)) {
+    syncDebug("[%d:%d] is unpromotable and can not campaign", pRaft->selfGroupId, pRaft->selfId);
+    return 0;
+  }
+  // if there is pending uncommitted config,cannot start election
+  if (syncRaftLogNumOfPendingConf(pRaft->log) > 0 && syncRaftHasUnappliedLog(pRaft->log)) {
+    syncWarn("[%d:%d] cannot syncRaftStartElection at term %" PRId64 " since there are still pending configuration changes to apply",
+      pRaft->selfGroupId, pRaft->selfId, pRaft->term);
+    return 0;
+  }
+
+  syncInfo("[%d:%d] is starting a new election at term %" PRId64 "", pRaft->selfGroupId, pRaft->selfId, pRaft->term);
+
+  if (pRaft->preVote) {
+    syncRaftStartElection(pRaft, SYNC_RAFT_CAMPAIGN_PRE_ELECTION);
+  } else {
+    syncRaftStartElection(pRaft, SYNC_RAFT_CAMPAIGN_ELECTION);
+  }
+
+  return 0;
+}

source/libs/sync/src/raft_handle_vote_message.c

@@ -0,0 +1,58 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <cli@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "syncInt.h"
+#include "raft.h"
+#include "raft_configuration.h"
+#include "raft_log.h"
+#include "raft_message.h"
+
+static bool canGrantVoteMessage(SSyncRaft* pRaft, const SSyncMessage* pMsg);
+
+int syncRaftHandleVoteMessage(SSyncRaft* pRaft, const SSyncMessage* pMsg) {
+  SSyncMessage* pRespMsg;
+  int voteIndex = syncRaftConfigurationIndexOfNode(pRaft, pMsg->from);
+  if (voteIndex == -1) {
+    return 0;
+  }
+  bool grant;
+  SyncIndex lastIndex = syncRaftLogLastIndex(pRaft->log);
+  SyncTerm lastTerm = syncRaftLogLastTerm(pRaft->log);
+
+  grant = canGrantVoteMessage(pRaft, pMsg);
+  pRespMsg = syncNewVoteRespMsg(pRaft->selfGroupId, pRaft->selfId, pMsg->vote.cType, !grant);
+  if (pRespMsg == NULL) {
+    return 0;
+  }
+  syncInfo("[%d:%d] [logterm: %" PRId64 ", index: %" PRId64 ", vote: %d] %s for %d"\    
+    "[logterm: %" PRId64 ", index: %" PRId64 "] at term %" PRId64 "",
+    pRaft->selfGroupId, pRaft->selfId, lastTerm, lastIndex, pRaft->voteFor,
+    grant ? "grant" : "reject",
+    pMsg->from, pMsg->vote.lastTerm, pMsg->vote.lastIndex, pRaft->term);
+
+  pRaft->io.send(pRespMsg, &(pRaft->cluster.nodeInfo[voteIndex]));
+  return 0;
+}
+
+static bool canGrantVoteMessage(SSyncRaft* pRaft, const SSyncMessage* pMsg) {
+  if (!(pRaft->voteFor == SYNC_NON_NODE_ID || pMsg->term > pRaft->term || pRaft->voteFor == pMsg->from)) {
+    return false;
+  }
+  if (!syncRaftLogIsUptodate(pRaft->log, pMsg->vote.lastIndex, pMsg->vote.lastTerm)) {
+    return false;
+  }
+
+  return true;
+}