diff --git a/src/client/src/tscSQLParser.c b/src/client/src/tscSQLParser.c
index ca6dfcee9efb03058ad2401ed57f2dce6f581d59..056314460dc18fd787a76a63124a55359dcdf5c2 100644
--- a/src/client/src/tscSQLParser.c
+++ b/src/client/src/tscSQLParser.c
@@ -4962,6 +4962,7 @@ static void setCreateDBOption(SCMCreateDbMsg* pMsg, SCreateDBInfo* pCreateDb) {
   pMsg->commitTime = htonl(pCreateDb->commitTime);
   pMsg->minRowsPerFileBlock = htonl(pCreateDb->minRowsPerBlock);
   pMsg->maxRowsPerFileBlock = htonl(pCreateDb->maxRowsPerBlock);
+  pMsg->fsyncPeriod = htonl(pCreateDb->fsyncPeriod);
   pMsg->compression = pCreateDb->compressionLevel;
   pMsg->walLevel = (char)pCreateDb->walLevel;
   pMsg->replications = pCreateDb->replica;
diff --git a/src/common/inc/tglobal.h b/src/common/inc/tglobal.h
index fbdfc84f5ffc26815c8318ff93f494af7801b6d9..235441e0a56579a32f9372686b2f64dbb55e3b48 100644
--- a/src/common/inc/tglobal.h
+++ b/src/common/inc/tglobal.h
@@ -78,6 +78,7 @@ extern int16_t tsCommitTime;  // seconds
 extern int32_t tsTimePrecision;
 extern int16_t tsCompression;
 extern int16_t tsWAL;
+extern int32_t tsFsyncPeriod;
 extern int32_t tsReplications;
 
 // balance
diff --git a/src/common/src/tglobal.c b/src/common/src/tglobal.c
index a8ef3f8e8d348b6b67fae6cc1dcd3f810b57e3ec..2bc0f3459c98525b42fd4323b74183223249cb39 100644
--- a/src/common/src/tglobal.c
+++ b/src/common/src/tglobal.c
@@ -110,6 +110,7 @@ int16_t tsCommitTime    = TSDB_DEFAULT_COMMIT_TIME;  // seconds
 int32_t tsTimePrecision = TSDB_DEFAULT_PRECISION;
 int16_t tsCompression   = TSDB_DEFAULT_COMP_LEVEL;
 int16_t tsWAL           = TSDB_DEFAULT_WAL_LEVEL;
+int32_t tsFsyncPeriod   = TSDB_DEFAULT_FSYNC_PERIOD;
 int32_t tsReplications  = TSDB_DEFAULT_DB_REPLICA_OPTION;
 int32_t tsMaxVgroupsPerDb  = 0;
 int32_t tsMaxTablePerVnode = TSDB_DEFAULT_TABLES;
@@ -702,6 +703,16 @@ static void doInitGlobalConfig() {
   cfg.unitType = TAOS_CFG_UTYPE_NONE;
   taosInitConfigOption(cfg);
 
+  cfg.option = "fsyncPeriod";
+  cfg.ptr = &tsFsyncPeriod;
+  cfg.valType = TAOS_CFG_VTYPE_INT32;
+  cfg.cfgType = TSDB_CFG_CTYPE_B_CONFIG | TSDB_CFG_CTYPE_B_SHOW;
+  cfg.minValue = TSDB_MIN_FSYNC_PERIOD;
+  cfg.maxValue = TSDB_MAX_FSYNC_PERIOD;
+  cfg.ptrLength = 0;
+  cfg.unitType = TAOS_CFG_UTYPE_NONE;
+  taosInitConfigOption(cfg);
+
   cfg.option = "replica";
   cfg.ptr = &tsReplications;
   cfg.valType = TAOS_CFG_VTYPE_INT32;
diff --git a/src/inc/taosdef.h b/src/inc/taosdef.h
index eb3c3abc840121043e7a36ef884f63f4c5caef57..b9cc35def5b5c9598f3a68b14ba4332652e1a45e 100644
--- a/src/inc/taosdef.h
+++ b/src/inc/taosdef.h
@@ -332,6 +332,10 @@ void tsDataSwap(void *pLeft, void *pRight, int32_t type, int32_t size);
 #define TSDB_MAX_WAL_LEVEL              2
 #define TSDB_DEFAULT_WAL_LEVEL          1
 
+#define TSDB_MIN_FSYNC_PERIOD           0
+#define TSDB_MAX_FSYNC_PERIOD           180000   // millisecond
+#define TSDB_DEFAULT_FSYNC_PERIOD       1000     // one second
+
 #define TSDB_MIN_DB_REPLICA_OPTION      1
 #define TSDB_MAX_DB_REPLICA_OPTION      3
 #define TSDB_DEFAULT_DB_REPLICA_OPTION  1
diff --git a/src/inc/taosmsg.h b/src/inc/taosmsg.h
index d96e9f68d5a33addb84ca08d286be7b88b62db30..dec3f60b649a9f852c4ca01925f9d27023badc13 100644
--- a/src/inc/taosmsg.h
+++ b/src/inc/taosmsg.h
@@ -515,6 +515,7 @@ typedef struct {
   int32_t  minRowsPerFileBlock;
   int32_t  maxRowsPerFileBlock;
   int32_t  commitTime;
+  int32_t  fsyncPeriod;
   uint8_t  precision;   // time resolution
   int8_t   compression;
   int8_t   walLevel;
@@ -608,6 +609,7 @@ typedef struct {
   int32_t  minRowsPerFileBlock;
   int32_t  maxRowsPerFileBlock;
   int32_t  commitTime;
+  int32_t  fsyncPeriod;
   int8_t   precision;
   int8_t   compression;
   int8_t   walLevel;
diff --git a/src/inc/twal.h b/src/inc/twal.h
index 8c2c3c69a1261ea0d86ac8d3b09c435a74b65085..4fdb7aa275d539ae63129e038c4ae50f59552e1f 100644
--- a/src/inc/twal.h
+++ b/src/inc/twal.h
@@ -35,6 +35,7 @@ typedef struct {
 
 typedef struct {
   int8_t    walLevel;  // wal level
+  int32_t   fsyncPeriod; // millisecond
   int8_t    wals;      // number of WAL files;
   int8_t    keep;      // keep the wal file when closed
 } SWalCfg;
diff --git a/src/kit/taosmigrate/taosmigrateVnodeCfg.c b/src/kit/taosmigrate/taosmigrateVnodeCfg.c
index b925fb10aa3bb59951726497b3c77a7a8d5ef142..1a1087aa9d3db424cfb8194e7611b86504339341 100644
--- a/src/kit/taosmigrate/taosmigrateVnodeCfg.c
+++ b/src/kit/taosmigrate/taosmigrateVnodeCfg.c
@@ -48,6 +48,7 @@ static int32_t saveVnodeCfg(SVnodeObj *pVnode, char* cfgFile)
   len += snprintf(content + len, maxLen - len, "  \"precision\": %d,\n", pVnode->tsdbCfg.precision);
   len += snprintf(content + len, maxLen - len, "  \"compression\": %d,\n", pVnode->tsdbCfg.compression);
   len += snprintf(content + len, maxLen - len, "  \"walLevel\": %d,\n", pVnode->walCfg.walLevel);
+  len += snprintf(content + len, maxLen - len, "  \"fsyncPeriod\": %d,\n", pVnode->walCfg.fsyncPeriod);
   len += snprintf(content + len, maxLen - len, "  \"replica\": %d,\n", pVnode->syncCfg.replica);
   len += snprintf(content + len, maxLen - len, "  \"wals\": %d,\n", pVnode->walCfg.wals);
   len += snprintf(content + len, maxLen - len, "  \"quorum\": %d,\n", pVnode->syncCfg.quorum);
@@ -212,6 +213,13 @@ static int32_t readVnodeCfg(SVnodeObj *pVnode, char* cfgFile)
   }
   pVnode->walCfg.walLevel = (int8_t) walLevel->valueint;
 
+  cJSON *fsyncPeriod = cJSON_GetObjectItem(root, "fsyncPeriod");
+  if (!fsyncPeriod || fsyncPeriod->type != cJSON_Number) {
+    printf("vgId:%d, failed to read vnode cfg, fsyncPeriod not found\n", pVnode->vgId);
+    goto PARSE_OVER;
+  }
+  pVnode->walCfg.fsyncPeriod = fsyncPeriod->valueint;
+
   cJSON *wals = cJSON_GetObjectItem(root, "wals");
   if (!wals || wals->type != cJSON_Number) {
     printf("vgId:%d, failed to read vnode cfg, wals not found\n", pVnode->vgId);
diff --git a/src/mnode/inc/mnodeDef.h b/src/mnode/inc/mnodeDef.h
index 90c2ff9e66d01a7fc89f4a1d8861025bc7668e7b..b7489a9fc56553c55ea4097a5df65b3a34f7b54b 100644
--- a/src/mnode/inc/mnodeDef.h
+++ b/src/mnode/inc/mnodeDef.h
@@ -160,6 +160,7 @@ typedef struct {
   int32_t minRowsPerFileBlock;
   int32_t maxRowsPerFileBlock;
   int32_t commitTime;
+  int32_t fsyncPeriod;
   int8_t  precision;
   int8_t  compression;
   int8_t  walLevel;
diff --git a/src/mnode/src/mnodeDb.c b/src/mnode/src/mnodeDb.c
index e8183e9089bd97cc40c20368cfcfa25ca450a368..6bb6f450a582d2036db063f3896b76ff4a01511f 100644
--- a/src/mnode/src/mnodeDb.c
+++ b/src/mnode/src/mnodeDb.c
@@ -287,14 +287,14 @@ static int32_t mnodeCheckDbCfg(SDbCfg *pCfg) {
     return TSDB_CODE_MND_INVALID_DB_OPTION;
   }
 
-  if (pCfg->replications < TSDB_MIN_DB_REPLICA_OPTION || pCfg->replications > TSDB_MAX_DB_REPLICA_OPTION) {
-    mError("invalid db option replications:%d valid range: [%d, %d]", pCfg->replications, TSDB_MIN_DB_REPLICA_OPTION,
-           TSDB_MAX_DB_REPLICA_OPTION);
+  if (pCfg->fsyncPeriod < TSDB_MIN_FSYNC_PERIOD || pCfg->fsyncPeriod > TSDB_MAX_FSYNC_PERIOD) {
+    mError("invalid db option fsyncPeriod:%d, valid range: [%d, %d]", pCfg->fsyncPeriod, TSDB_MIN_FSYNC_PERIOD, TSDB_MAX_FSYNC_PERIOD);
     return TSDB_CODE_MND_INVALID_DB_OPTION;
   }
 
-  if (pCfg->walLevel < TSDB_MIN_WAL_LEVEL) {
-    mError("invalid db option walLevel:%d must be greater than 0", pCfg->walLevel);
+  if (pCfg->replications < TSDB_MIN_DB_REPLICA_OPTION || pCfg->replications > TSDB_MAX_DB_REPLICA_OPTION) {
+    mError("invalid db option replications:%d valid range: [%d, %d]", pCfg->replications, TSDB_MIN_DB_REPLICA_OPTION,
+           TSDB_MAX_DB_REPLICA_OPTION);
     return TSDB_CODE_MND_INVALID_DB_OPTION;
   }
 
@@ -318,6 +318,7 @@ static void mnodeSetDefaultDbCfg(SDbCfg *pCfg) {
   if (pCfg->daysToKeep2 < 0) pCfg->daysToKeep2 = pCfg->daysToKeep;
   if (pCfg->minRowsPerFileBlock < 0) pCfg->minRowsPerFileBlock = tsMinRowsInFileBlock;
   if (pCfg->maxRowsPerFileBlock < 0) pCfg->maxRowsPerFileBlock = tsMaxRowsInFileBlock;
+  if (pCfg->fsyncPeriod <0) pCfg->fsyncPeriod = tsFsyncPeriod;
   if (pCfg->commitTime < 0) pCfg->commitTime = tsCommitTime;
   if (pCfg->precision < 0) pCfg->precision = tsTimePrecision;
   if (pCfg->compression < 0) pCfg->compression = tsCompression;
@@ -367,6 +368,7 @@ static int32_t mnodeCreateDb(SAcctObj *pAcct, SCMCreateDbMsg *pCreate, void *pMs
     .daysToKeep2         = pCreate->daysToKeep2,
     .minRowsPerFileBlock = pCreate->minRowsPerFileBlock,
     .maxRowsPerFileBlock = pCreate->maxRowsPerFileBlock,
+    .fsyncPeriod         = pCreate->fsyncPeriod,
     .commitTime          = pCreate->commitTime,
     .precision           = pCreate->precision,
     .compression         = pCreate->compression,
@@ -682,6 +684,10 @@ static int32_t mnodeRetrieveDbs(SShowObj *pShow, char *data, int32_t rows, void
       *(int8_t *)pWrite = pDb->cfg.walLevel;
       cols++;
 
+      pWrite = data + pShow->offset[cols] * rows + pShow->bytes[cols] * numOfRows;
+      *(int32_t *)pWrite = pDb->cfg.fsyncPeriod;
+      cols++;
+
       pWrite = data + pShow->offset[cols] * rows + pShow->bytes[cols] * numOfRows;
       *(int8_t *)pWrite = pDb->cfg.compression;
       cols++;
@@ -785,6 +791,7 @@ static SDbCfg mnodeGetAlterDbOption(SDbObj *pDb, SCMAlterDbMsg *pAlter) {
   int32_t minRows        = htonl(pAlter->minRowsPerFileBlock);
   int32_t maxRows        = htonl(pAlter->maxRowsPerFileBlock);
   int32_t commitTime     = htonl(pAlter->commitTime);
+  int32_t fsyncPeriod    = htonl(pAlter->fsyncPeriod);
   int8_t  compression    = pAlter->compression;
   int8_t  walLevel       = pAlter->walLevel;
   int8_t  replications   = pAlter->replications;
@@ -861,6 +868,11 @@ static SDbCfg mnodeGetAlterDbOption(SDbObj *pDb, SCMAlterDbMsg *pAlter) {
     terrno = TSDB_CODE_MND_INVALID_DB_OPTION;
   }
 
+  if (fsyncPeriod >= 0 && fsyncPeriod != pDb->cfg.fsyncPeriod) {
+    mError("db:%s, can't alter fsyncPeriod option", pDb->name);
+    terrno = TSDB_CODE_MND_INVALID_DB_OPTION;
+  }
+
   if (replications > 0 && replications != pDb->cfg.replications) {
     mDebug("db:%s, replications:%d change to %d", pDb->name, pDb->cfg.replications, replications);
     newCfg.replications = replications;
diff --git a/src/mnode/src/mnodeSdb.c b/src/mnode/src/mnodeSdb.c
index d3192a2460716ee31881715c4146de4330c8ca0c..1c11d25baac01e5914f6bd6cd954268dc5c401a8 100644
--- a/src/mnode/src/mnodeSdb.c
+++ b/src/mnode/src/mnodeSdb.c
@@ -170,7 +170,7 @@ static void *sdbGetTableFromId(int32_t tableId) {
 }
 
 static int32_t sdbInitWal() {
-  SWalCfg walCfg = {.walLevel = 2, .wals = 2, .keep = 1};
+  SWalCfg walCfg = {.walLevel = 2, .wals = 2, .keep = 1, .fsyncPeriod = 0};
   char temp[TSDB_FILENAME_LEN];
   sprintf(temp, "%s/wal", tsMnodeDir);
   tsSdbObj.wal = walOpen(temp, &walCfg);
diff --git a/src/mnode/src/mnodeVgroup.c b/src/mnode/src/mnodeVgroup.c
index 5658f4bcbf17ac2f6f0fa7981f6abf3c170dada7..40d896606c5c27a33c62be90cddad17d29a8249c 100644
--- a/src/mnode/src/mnodeVgroup.c
+++ b/src/mnode/src/mnodeVgroup.c
@@ -755,6 +755,7 @@ SMDCreateVnodeMsg *mnodeBuildCreateVnodeMsg(SVgObj *pVgroup) {
   pCfg->daysToKeep2         = htonl(pDb->cfg.daysToKeep2);  
   pCfg->minRowsPerFileBlock = htonl(pDb->cfg.minRowsPerFileBlock);
   pCfg->maxRowsPerFileBlock = htonl(pDb->cfg.maxRowsPerFileBlock);
+  pCfg->fsyncPeriod         = htonl(pDb->cfg.fsyncPeriod);
   pCfg->commitTime          = htonl(pDb->cfg.commitTime);
   pCfg->precision           = pDb->cfg.precision;
   pCfg->compression         = pDb->cfg.compression;
diff --git a/src/query/inc/qsqlparser.h b/src/query/inc/qsqlparser.h
index 704f3e74187ef1686be1ec7babd30b3fe2b2b1e1..9e8706cf9220c2a0d0e55fb0a8f1028a876da9d4 100644
--- a/src/query/inc/qsqlparser.h
+++ b/src/query/inc/qsqlparser.h
@@ -116,6 +116,7 @@ typedef struct SCreateDBInfo {
   int32_t   daysPerFile;
   int32_t   minRowsPerBlock;
   int32_t   maxRowsPerBlock;
+  int32_t   fsyncPeriod;
   int64_t   commitTime;
   int32_t   walLevel;
   int32_t   compressionLevel;
diff --git a/src/query/inc/sql.y b/src/query/inc/sql.y
index b4ea1254b7094a8f9a3f83d55da63d990dc21b09..aa6f794d5b75c7e49451c50d038dff3aa4a88abc 100644
--- a/src/query/inc/sql.y
+++ b/src/query/inc/sql.y
@@ -221,6 +221,7 @@ maxrows(Y) ::= MAXROWS INTEGER(X).            { Y = X; }
 blocks(Y)  ::= BLOCKS INTEGER(X).             { Y = X; }
 ctime(Y)   ::= CTIME INTEGER(X).              { Y = X; }
 wal(Y)     ::= WAL INTEGER(X).                { Y = X; }
+fsync(Y)   ::= FSYNC INTEGER(X).              { Y = X; }
 comp(Y)    ::= COMP INTEGER(X).               { Y = X; }
 prec(Y)    ::= PRECISION STRING(X).           { Y = X; }
 
@@ -233,6 +234,7 @@ db_optr(Y) ::= db_optr(Z) replica(X).        { Y = Z; Y.replica = strtol(X.z, NU
 db_optr(Y) ::= db_optr(Z) days(X).           { Y = Z; Y.daysPerFile = strtol(X.z, NULL, 10); }
 db_optr(Y) ::= db_optr(Z) minrows(X).        { Y = Z; Y.minRowsPerBlock = strtod(X.z, NULL); }
 db_optr(Y) ::= db_optr(Z) maxrows(X).        { Y = Z; Y.maxRowsPerBlock = strtod(X.z, NULL); }
+db_optr(Y) ::= db_optr(Z) fsync(X).          { Y = Z; Y.fsyncPeriod = strtod(X.z, NULL); }
 db_optr(Y) ::= db_optr(Z) blocks(X).         { Y = Z; Y.numOfBlocks = strtol(X.z, NULL, 10); }
 db_optr(Y) ::= db_optr(Z) ctime(X).          { Y = Z; Y.commitTime = strtol(X.z, NULL, 10); }
 db_optr(Y) ::= db_optr(Z) wal(X).            { Y = Z; Y.walLevel = strtol(X.z, NULL, 10); }
@@ -249,6 +251,7 @@ alter_db_optr(Y) ::= alter_db_optr(Z) keep(X).        { Y = Z; Y.keep = X; }
 alter_db_optr(Y) ::= alter_db_optr(Z) blocks(X).      { Y = Z; Y.numOfBlocks = strtol(X.z, NULL, 10); }
 alter_db_optr(Y) ::= alter_db_optr(Z) comp(X).        { Y = Z; Y.compressionLevel = strtol(X.z, NULL, 10); }
 alter_db_optr(Y) ::= alter_db_optr(Z) wal(X).         { Y = Z; Y.walLevel = strtol(X.z, NULL, 10); }
+alter_db_optr(Y) ::= alter_db_optr(Z) fsync(X).       { Y = Z; Y.fsyncPeriod = strtod(X.z, NULL, 10); }
 
 %type typename {TAOS_FIELD}
 typename(A) ::= ids(X). { 
diff --git a/src/query/src/qparserImpl.c b/src/query/src/qparserImpl.c
index d4ac540d2fc725488bae83fb8890840f6ddb59a9..ecc11f8f4d76769d0d52b03f44813349aed29adc 100644
--- a/src/query/src/qparserImpl.c
+++ b/src/query/src/qparserImpl.c
@@ -896,6 +896,7 @@ void setDefaultCreateDbOption(SCreateDBInfo *pDBInfo) {
   pDBInfo->compressionLevel = -1;
 
   pDBInfo->walLevel = -1;
+  pDBInfo->fsyncPeriod = -1;
   pDBInfo->commitTime = -1;
   pDBInfo->maxTablesPerVnode = -1;
 
diff --git a/src/vnode/src/vnodeMain.c b/src/vnode/src/vnodeMain.c
index 695a55d47638c1b0e05651be24d2480b9d5b975c..e46e400c15c77a4a4f9c1a74e9acf0a09dd9a10d 100644
--- a/src/vnode/src/vnodeMain.c
+++ b/src/vnode/src/vnodeMain.c
@@ -69,6 +69,7 @@ int32_t vnodeInitResources() {
 }
 
 void vnodeCleanupResources() {
+
   if (tsDnodeVnodesHash != NULL) {
     taosHashCleanup(tsDnodeVnodesHash);
     tsDnodeVnodesHash = NULL;
@@ -137,7 +138,7 @@ int32_t vnodeCreate(SMDCreateVnodeMsg *pVnodeCfg) {
     return TSDB_CODE_VND_INIT_FAILED;
   }
 
-  vInfo("vgId:%d, vnode is created, clog:%d", pVnodeCfg->cfg.vgId, pVnodeCfg->cfg.walLevel);
+  vInfo("vgId:%d, vnode is created, walLevel:%d fsyncPeriod:%d", pVnodeCfg->cfg.vgId, pVnodeCfg->cfg.walLevel, pVnodeCfg->cfg.fsyncPeriod);
   code = vnodeOpen(pVnodeCfg->cfg.vgId, rootDir);
 
   return code;
@@ -618,6 +619,7 @@ static int32_t vnodeSaveCfg(SMDCreateVnodeMsg *pVnodeCfg) {
   len += snprintf(content + len, maxLen - len, "  \"precision\": %d,\n", pVnodeCfg->cfg.precision);
   len += snprintf(content + len, maxLen - len, "  \"compression\": %d,\n", pVnodeCfg->cfg.compression);
   len += snprintf(content + len, maxLen - len, "  \"walLevel\": %d,\n", pVnodeCfg->cfg.walLevel);
+  len += snprintf(content + len, maxLen - len, "  \"fsyncPeriod\": %d,\n", pVnodeCfg->cfg.fsyncPeriod);
   len += snprintf(content + len, maxLen - len, "  \"replica\": %d,\n", pVnodeCfg->cfg.replications);
   len += snprintf(content + len, maxLen - len, "  \"wals\": %d,\n", pVnodeCfg->cfg.wals);
   len += snprintf(content + len, maxLen - len, "  \"quorum\": %d,\n", pVnodeCfg->cfg.quorum);
@@ -782,6 +784,13 @@ static int32_t vnodeReadCfg(SVnodeObj *pVnode) {
   }
   pVnode->walCfg.walLevel = (int8_t) walLevel->valueint;
 
+  cJSON *fsyncPeriod = cJSON_GetObjectItem(root, "fsyncPeriod");
+  if (!walLevel || walLevel->type != cJSON_Number) {
+    vError("vgId:%d, failed to read vnode cfg, fsyncPeriod not found", pVnode->vgId);
+    goto PARSE_OVER;
+  }
+  pVnode->walCfg.fsyncPeriod = fsyncPeriod->valueint;
+
   cJSON *wals = cJSON_GetObjectItem(root, "wals");
   if (!wals || wals->type != cJSON_Number) {
     vError("vgId:%d, failed to read vnode cfg, wals not found", pVnode->vgId);
diff --git a/src/wal/src/walMain.c b/src/wal/src/walMain.c
index 94a0fdc956577d9cc2e35f6bd780dcb6b8b33950..45bacc317fa14d076f404f34b2d0c81dbb38b648 100644
--- a/src/wal/src/walMain.c
+++ b/src/wal/src/walMain.c
@@ -25,6 +25,7 @@
 #include "tlog.h"
 #include "tchecksum.h"
 #include "tutil.h"
+#include "ttimer.h"
 #include "taoserror.h"
 #include "twal.h"
 #include "tqueue.h"
@@ -44,6 +45,9 @@ typedef struct {
   int      fd;
   int      keep;
   int      level;
+  int32_t  fsyncPeriod;
+  void    *timer;
+  void    *signature;
   int      max;  // maximum number of wal files
   uint32_t id;   // increase continuously
   int      num;  // number of wal files
@@ -52,10 +56,20 @@ typedef struct {
   pthread_mutex_t mutex;
 } SWal;
 
+static void    *walTmrCtrl = NULL;
+static int     tsWalNum = 0;
+static pthread_once_t walModuleInit = PTHREAD_ONCE_INIT;
 static uint32_t walSignature = 0xFAFBFDFE;
-static int walHandleExistingFiles(const char *path);
-static int walRestoreWalFile(SWal *pWal, void *pVnode, FWalWrite writeFp);
-static int walRemoveWalFiles(const char *path);
+static int  walHandleExistingFiles(const char *path);
+static int  walRestoreWalFile(SWal *pWal, void *pVnode, FWalWrite writeFp);
+static int  walRemoveWalFiles(const char *path);
+static void walProcessFsyncTimer(void *param, void *tmrId);
+static void walRelease(SWal *pWal);
+
+static void walModuleInitFunc() {
+  walTmrCtrl = taosTmrInit(1000, 100, 300000, "WAL");
+  if (walTmrCtrl == NULL) walModuleInit = PTHREAD_ONCE_INIT;
+}
 
 void *walOpen(const char *path, const SWalCfg *pCfg) {
   SWal *pWal = calloc(sizeof(SWal), 1);
@@ -64,20 +78,38 @@ void *walOpen(const char *path, const SWalCfg *pCfg) {
     return NULL;
   }
 
+  pthread_once(&walModuleInit, walModuleInitFunc);
+  if (walTmrCtrl == NULL) {
+    free(pWal);
+    terrno = TAOS_SYSTEM_ERROR(errno);
+    return NULL;
+  }
+
+  atomic_add_fetch_32(&tsWalNum, 1);    
   pWal->fd = -1;
   pWal->max = pCfg->wals;
   pWal->id = 0;
   pWal->num = 0;
   pWal->level = pCfg->walLevel;
   pWal->keep = pCfg->keep;
+  pWal->fsyncPeriod = pCfg->fsyncPeriod;
+  pWal->signature = pWal;
   tstrncpy(pWal->path, path, sizeof(pWal->path));
   pthread_mutex_init(&pWal->mutex, NULL);
 
+  if (pWal->fsyncPeriod > 0  && pWal->level == TAOS_WAL_FSYNC) {
+    pWal->timer = taosTmrStart(walProcessFsyncTimer, pWal->fsyncPeriod, pWal, walTmrCtrl);
+    if (pWal->timer == NULL) {
+      terrno = TAOS_SYSTEM_ERROR(errno);
+      walRelease(pWal);
+      return NULL;
+    }
+  }
+
   if (tmkdir(path, 0755) != 0) {
     terrno = TAOS_SYSTEM_ERROR(errno);
     wError("wal:%s, failed to create directory(%s)", path, strerror(errno));
-    pthread_mutex_destroy(&pWal->mutex);
-    free(pWal);
+    walRelease(pWal);
     pWal = NULL;
   }
      
@@ -89,12 +121,11 @@ void *walOpen(const char *path, const SWalCfg *pCfg) {
   if (pWal && pWal->fd <0) {
     terrno = TAOS_SYSTEM_ERROR(errno);
     wError("wal:%s, failed to open(%s)", path, strerror(errno));
-    pthread_mutex_destroy(&pWal->mutex);
-    free(pWal);
+    walRelease(pWal);
     pWal = NULL;
   } 
 
-  if (pWal) wDebug("wal:%s, it is open, level:%d", path, pWal->level);
+  if (pWal) wDebug("wal:%s, it is open, level:%d fsyncPeriod:%d", path, pWal->level, pWal->fsyncPeriod);
   return pWal;
 }
 
@@ -102,7 +133,8 @@ void walClose(void *handle) {
   if (handle == NULL) return;
   
   SWal *pWal = handle;  
-  close(pWal->fd);
+  tclose(pWal->fd);
+  if (pWal->timer) taosTmrStopA(&pWal->timer);
 
   if (pWal->keep == 0) {
     // remove all files in the directory
@@ -118,9 +150,7 @@ void walClose(void *handle) {
     wDebug("wal:%s, it is closed and kept", pWal->name);
   }
 
-  pthread_mutex_destroy(&pWal->mutex);
-
-  free(pWal);
+  walRelease(pWal);
 }
 
 int walRenew(void *handle) {
@@ -194,9 +224,9 @@ int walWrite(void *handle, SWalHead *pHead) {
 void walFsync(void *handle) {
 
   SWal *pWal = handle;
-  if (pWal == NULL) return;
+  if (pWal == NULL || pWal->level != TAOS_WAL_FSYNC || pWal->fd < 0) return;
 
-  if (pWal->level == TAOS_WAL_FSYNC && pWal->fd >=0) {
+  if (pWal->fsyncPeriod == 0) {
     if (fsync(pWal->fd) < 0) {
       wError("wal:%s, fsync failed(%s)", pWal->name, strerror(errno));
     }
@@ -303,6 +333,19 @@ int walGetWalFile(void *handle, char *name, uint32_t *index) {
   return code;
 }  
 
+static void walRelease(SWal *pWal) {
+
+  pthread_mutex_destroy(&pWal->mutex);
+  pWal->signature = NULL;
+  free(pWal);
+
+  if (atomic_sub_fetch_32(&tsWalNum, 1) == 0) {
+    if (walTmrCtrl) taosTmrCleanUp(walTmrCtrl);
+    walTmrCtrl = NULL;
+    walModuleInit = PTHREAD_ONCE_INIT;
+  }
+}
+
 static int walRestoreWalFile(SWal *pWal, void *pVnode, FWalWrite writeFp) {
   char *name = pWal->name;
 
@@ -433,3 +476,15 @@ static int walRemoveWalFiles(const char *path) {
   return terrno;
 }
 
+static void walProcessFsyncTimer(void *param, void *tmrId) {
+  SWal *pWal = param;
+
+  if (pWal->signature != pWal) return;
+  if (pWal->fd < 0) return;
+
+  if (fsync(pWal->fd) < 0) {
+    wError("wal:%s, fsync failed(%s)", pWal->name, strerror(errno));
+  }
+  
+  pWal->timer = taosTmrStart(walProcessFsyncTimer, pWal->fsyncPeriod, pWal, walTmrCtrl);
+}