fix(cluster): get monitor info in multi-process mode

include/libs/monitor/monitor.h

@@ -78,6 +78,9 @@ typedef struct {
 typedef struct {
   float        uptime;  // day
   int8_t       has_mnode;
+  int8_t       has_qnode;
+  int8_t       has_snode;
+  int8_t       has_bnode;
   SMonDiskDesc logdir;
   SMonDiskDesc tempdir;
 } SMonDnodeInfo;
@@ -134,8 +137,8 @@ typedef struct {
 
 typedef struct {
   int32_t expire_time;
-  int32_t timeseries_used;
-  int32_t timeseries_total;
+  int64_t timeseries_used;
+  int64_t timeseries_total;
 } SMonGrantInfo;
 
 typedef struct {

include/os/osSysinfo.h

@@ -39,7 +39,7 @@ int32_t taosGetEmail(char *email, int32_t maxLen);
 int32_t taosGetOsReleaseName(char *releaseName, int32_t maxLen);
 int32_t taosGetCpuInfo(char *cpuModel, int32_t maxLen, float *numOfCores);
 int32_t taosGetCpuCores(float *numOfCores);
-int32_t taosGetCpuUsage(double *cpu_system, double *cpu_engine);
+void    taosGetCpuUsage(double *cpu_system, double *cpu_engine);
 int32_t taosGetTotalMemory(int64_t *totalKB);
 int32_t taosGetProcMemory(int64_t *usedKB);
 int32_t taosGetSysMemory(int64_t *usedKB);

source/dnode/mgmt/dm/dmMonitor.c

@@ -25,11 +25,10 @@ static void dmGetMonitorBasicInfo(SDnode *pDnode, SMonBasicInfo *pInfo) {
 
 static void dmGetMonitorDnodeInfo(SDnode *pDnode, SMonDnodeInfo *pInfo) {
   pInfo->uptime = (taosGetTimestampMs() - pDnode->rebootTime) / (86400000.0f);
-  SMgmtWrapper *pWrapper = dndAcquireWrapper(pDnode, MNODE);
-  if (pWrapper != NULL) {
-    pInfo->has_mnode = pWrapper->required;
-    dndReleaseWrapper(pWrapper);
-  }
+  pInfo->has_mnode = pDnode->wrappers[MNODE].required;
+  pInfo->has_qnode = pDnode->wrappers[QNODE].required;
+  pInfo->has_snode = pDnode->wrappers[SNODE].required;
+  pInfo->has_bnode = pDnode->wrappers[BNODE].required;
   tstrncpy(pInfo->logdir.name, tsLogDir, sizeof(pInfo->logdir.name));
   pInfo->logdir.size = tsLogSpace.size;
   tstrncpy(pInfo->tempdir.name, tsTempDir, sizeof(pInfo->tempdir.name));
@@ -65,7 +64,7 @@ void dmSendMonitorReport(SDnode *pDnode) {
   bool getFromAPI = !tsMultiProcess;
   pWrapper = &pDnode->wrappers[MNODE];
   if (getFromAPI) {
-    if (dndMarkWrapper(pWrapper) != 0) {
+    if (dndMarkWrapper(pWrapper) == 0) {
       mmGetMonitorInfo(pWrapper, &mmInfo);
       dndReleaseWrapper(pWrapper);
     }
@@ -82,7 +81,7 @@ void dmSendMonitorReport(SDnode *pDnode) {
 
   pWrapper = &pDnode->wrappers[VNODES];
   if (getFromAPI) {
-    if (dndMarkWrapper(pWrapper) != 0) {
+    if (dndMarkWrapper(pWrapper) == 0) {
       vmGetMonitorInfo(pWrapper, &vmInfo);
       dndReleaseWrapper(pWrapper);
     }
@@ -99,7 +98,7 @@ void dmSendMonitorReport(SDnode *pDnode) {
 
   pWrapper = &pDnode->wrappers[QNODE];
   if (getFromAPI) {
-    if (dndMarkWrapper(pWrapper) != 0) {
+    if (dndMarkWrapper(pWrapper) == 0) {
       qmGetMonitorInfo(pWrapper, &qmInfo);
       dndReleaseWrapper(pWrapper);
     }
@@ -116,7 +115,7 @@ void dmSendMonitorReport(SDnode *pDnode) {
 
   pWrapper = &pDnode->wrappers[SNODE];
   if (getFromAPI) {
-    if (dndMarkWrapper(pWrapper) != 0) {
+    if (dndMarkWrapper(pWrapper) == 0) {
       smGetMonitorInfo(pWrapper, &smInfo);
       dndReleaseWrapper(pWrapper);
     }
@@ -133,7 +132,7 @@ void dmSendMonitorReport(SDnode *pDnode) {
 
   pWrapper = &pDnode->wrappers[BNODE];
   if (getFromAPI) {
-    if (dndMarkWrapper(pWrapper) != 0) {
+    if (dndMarkWrapper(pWrapper) == 0) {
       bmGetMonitorInfo(pWrapper, &bmInfo);
       dndReleaseWrapper(pWrapper);
     }

source/dnode/mgmt/qm/qmWorker.c

@@ -32,7 +32,7 @@ static void qmProcessMonitorQueue(SQueueInfo *pInfo, SNodeMsg *pMsg) {
   SRpcMsg *pRpc = &pMsg->rpcMsg;
   int32_t  code = -1;
 
-  if (pMsg->rpcMsg.msgType == TDMT_MON_SM_INFO) {
+  if (pMsg->rpcMsg.msgType == TDMT_MON_QM_INFO) {
     code = qmProcessGetMonQmInfoReq(pMgmt->pWrapper, pMsg);
   } else {
     terrno = TSDB_CODE_MSG_NOT_PROCESSED;

source/dnode/mnode/impl/src/mndDnode.c

@@ -330,9 +330,11 @@ static int32_t mndProcessStatusReq(SNodeMsg *pReq) {
 
     SVgObj *pVgroup = mndAcquireVgroup(pMnode, pVload->vgId);
     if (pVgroup != NULL) {
+      mInfo("------>0 vgId:%d, numOfTimeSeries:%" PRId64, pVgroup->vgId, pVload->numOfTimeSeries);
       if (pVload->role == TAOS_SYNC_STATE_LEADER) {
         pVgroup->numOfTables = pVload->numOfTables;
         pVgroup->numOfTimeSeries = pVload->numOfTimeSeries;
+        mInfo("------>1 vgId:%d, numOfTimeSeries:%" PRId64, pVgroup->vgId, pVgroup->numOfTimeSeries);
         pVgroup->totalStorage = pVload->totalStorage;
         pVgroup->compStorage = pVload->compStorage;
         pVgroup->pointsWritten = pVload->pointsWritten;

source/libs/monitor/src/monMain.c

@@ -375,6 +375,9 @@ static void monGenDnodeJson(SMonInfo *pMonitor) {
   tjsonAddDoubleToObject(pJson, "vnodes_num", pStat->totalVnodes);
   tjsonAddDoubleToObject(pJson, "masters", pStat->masterNum);
   tjsonAddDoubleToObject(pJson, "has_mnode", pInfo->has_mnode);
+  tjsonAddDoubleToObject(pJson, "has_qnode", pInfo->has_qnode);
+  tjsonAddDoubleToObject(pJson, "has_snode", pInfo->has_snode);
+  tjsonAddDoubleToObject(pJson, "has_bnode", pInfo->has_bnode);
 }
 
 static void monGenDiskJson(SMonInfo *pMonitor) {

source/libs/monitor/src/monMsg.c

@@ -194,9 +194,9 @@ int32_t tDecodeSMonVgroupInfo(SCoder *decoder, SMonVgroupInfo *pInfo) {
     if (tDecodeCStrTo(decoder, desc.database_name) < 0) return -1;
     if (tDecodeCStrTo(decoder, desc.status) < 0) return -1;
     for (int32_t j = 0; j < TSDB_MAX_REPLICA; ++j) {
-      SMonVnodeDesc vdesc = {0};
-      if (tDecodeI32(decoder, &vdesc.dnode_id) < 0) return -1;
-      if (tDecodeCStrTo(decoder, vdesc.vnode_role) < 0) return -1;
+      SMonVnodeDesc *pVDesc = &desc.vnodes[j];
+      if (tDecodeI32(decoder, &pVDesc->dnode_id) < 0) return -1;
+      if (tDecodeCStrTo(decoder, pVDesc->vnode_role) < 0) return -1;
     }
     taosArrayPush(pInfo->vgroups, &desc);
   }
@@ -205,15 +205,15 @@ int32_t tDecodeSMonVgroupInfo(SCoder *decoder, SMonVgroupInfo *pInfo) {
 
 int32_t tEncodeSMonGrantInfo(SCoder *encoder, const SMonGrantInfo *pInfo) {
   if (tEncodeI32(encoder, pInfo->expire_time) < 0) return -1;
-  if (tEncodeI32(encoder, pInfo->timeseries_used) < 0) return -1;
-  if (tEncodeI32(encoder, pInfo->timeseries_total) < 0) return -1;
+  if (tEncodeI64(encoder, pInfo->timeseries_used) < 0) return -1;
+  if (tEncodeI64(encoder, pInfo->timeseries_total) < 0) return -1;
   return 0;
 }
 
 int32_t tDecodeSMonGrantInfo(SCoder *decoder, SMonGrantInfo *pInfo) {
   if (tDecodeI32(decoder, &pInfo->expire_time) < 0) return -1;
-  if (tDecodeI32(decoder, &pInfo->timeseries_used) < 0) return -1;
-  if (tDecodeI32(decoder, &pInfo->timeseries_total) < 0) return -1;
+  if (tDecodeI64(decoder, &pInfo->timeseries_used) < 0) return -1;
+  if (tDecodeI64(decoder, &pInfo->timeseries_total) < 0) return -1;
   return 0;
 }
 

source/os/src/osSysinfo.c

@@ -369,53 +369,33 @@ int32_t taosGetCpuCores(float *numOfCores) {
 #endif
 }
 
-int32_t taosGetCpuUsage(double *cpu_system, double *cpu_engine) {
-#if defined(_TD_WINDOWS_64) || defined(_TD_WINDOWS_32)
-  *cpu_system = 0;
-  *cpu_engine = 0;
-  return 0;
-#elif defined(_TD_DARWIN_64)
+void taosGetCpuUsage(double *cpu_system, double *cpu_engine) {
+  static int64_t lastSysUsed = 0;
+  static int64_t lastSysTotal = 0;
+  static int64_t lastProcTotal = 0;
+  static int64_t curSysUsed = 0;
+  static int64_t curSysTotal = 0;
+  static int64_t curProcTotal = 0;
+
   *cpu_system = 0;
   *cpu_engine = 0;
-  return 0;
-#else
-  static uint64_t lastSysUsed = 0;
-  static uint64_t lastSysTotal = 0;
-  static uint64_t lastProcTotal = 0;
-
-  SysCpuInfo  sysCpu;
-  ProcCpuInfo procCpu;
-  if (taosGetSysCpuInfo(&sysCpu) != 0) {
-    return -1;
-  }
-  if (taosGetProcCpuInfo(&procCpu) != 0) {
-    return -1;
-  }
 
-  uint64_t curSysUsed = sysCpu.user + sysCpu.nice + sysCpu.system;
-  uint64_t curSysTotal = curSysUsed + sysCpu.idle;
-  uint64_t curProcTotal = procCpu.utime + procCpu.stime + procCpu.cutime + procCpu.cstime;
+  SysCpuInfo  sysCpu = {0};
+  ProcCpuInfo procCpu = {0};
+  if (taosGetSysCpuInfo(&sysCpu) == 0 && taosGetProcCpuInfo(&procCpu) == 0) {
+    curSysUsed = sysCpu.user + sysCpu.nice + sysCpu.system;
+    curSysTotal = curSysUsed + sysCpu.idle;
+    curProcTotal = procCpu.utime + procCpu.stime + procCpu.cutime + procCpu.cstime;
 
-  if (lastSysUsed == 0 || lastSysTotal == 0 || lastProcTotal == 0) {
-    lastSysUsed = curSysUsed > 1 ? curSysUsed : 1;
-    lastSysTotal = curSysTotal > 1 ? curSysTotal : 1;
-    lastProcTotal = curProcTotal > 1 ? curProcTotal : 1;
-    return -1;
-  }
+    if (curSysTotal > lastSysTotal && curSysUsed >= lastSysUsed && curProcTotal >= lastProcTotal) {
+      *cpu_engine = (curSysUsed - lastSysUsed) / (double)(curSysTotal - lastSysTotal) * 100;
+      *cpu_system = (curProcTotal - lastProcTotal) / (double)(curSysTotal - lastSysTotal) * 100;
+    }
 
-  if (curSysTotal == lastSysTotal) {
-    return -1;
+    lastSysUsed = curSysUsed;
+    lastSysTotal = curSysTotal;
+    lastProcTotal = curProcTotal;
   }
-
-  *cpu_engine = (curSysUsed - lastSysUsed) / (double)(curSysTotal - lastSysTotal) * 100;
-  *cpu_system = (curProcTotal - lastProcTotal) / (double)(curSysTotal - lastSysTotal) * 100;
-
-  lastSysUsed = curSysUsed;
-  lastSysTotal = curSysTotal;
-  lastProcTotal = curProcTotal;
-
-  return 0;
-#endif
 }
 
 int32_t taosGetTotalMemory(int64_t *totalKB) {
@@ -618,7 +598,6 @@ void taosGetProcIODelta(int64_t *rchars, int64_t *wchars, int64_t *read_bytes, i
   static int64_t last_wchars = 0;
   static int64_t last_read_bytes = 0;
   static int64_t last_write_bytes = 0;
-
   static int64_t cur_rchars = 0;
   static int64_t cur_wchars = 0;
   static int64_t cur_read_bytes = 0;
@@ -632,6 +611,11 @@ void taosGetProcIODelta(int64_t *rchars, int64_t *wchars, int64_t *read_bytes, i
     last_wchars = cur_wchars;
     last_read_bytes = cur_read_bytes;
     last_write_bytes = cur_write_bytes;
+  } else {
+    *rchars = 0;
+    *wchars = 0;
+    *read_bytes = 0;
+    *write_bytes = 0;
   }
 }
 
@@ -693,7 +677,6 @@ int32_t taosGetCardInfo(int64_t *receive_bytes, int64_t *transmit_bytes) {
 void taosGetCardInfoDelta(int64_t *receive_bytes, int64_t *transmit_bytes) {
   static int64_t last_receive_bytes = 0;
   static int64_t last_transmit_bytes = 0;
-
   static int64_t cur_receive_bytes = 0;
   static int64_t cur_transmit_bytes = 0;
   if (taosGetCardInfo(&cur_receive_bytes, &cur_transmit_bytes) == 0) {
@@ -701,6 +684,9 @@ void taosGetCardInfoDelta(int64_t *receive_bytes, int64_t *transmit_bytes) {
     *transmit_bytes = cur_transmit_bytes - last_transmit_bytes;
     last_receive_bytes = cur_receive_bytes;
     last_transmit_bytes = cur_transmit_bytes;
+  } else {
+    *receive_bytes = 0;
+    *transmit_bytes = 0;
   }
 }
 

tests/script/tmp/monitor.sim

+system sh/stop_dnodes.sh
+system sh/deploy.sh -n dnode1 -i 1
+system sh/cfg.sh -n dnode1 -c monitorfqdn -v localhost
+system sh/cfg.sh -n dnode1 -c monitorport -v 80
+system sh/cfg.sh -n dnode1 -c monitorInterval -v 1
+system sh/cfg.sh -n dnode1 -c monitorComp -v 1
+#system sh/cfg.sh -n dnode1 -c supportVnodes -v 128
+
+system sh/exec.sh -n dnode1 -s start
+sql connect
+
+print =============== show dnodes
+sleep 2000
+sql create database db vgroups 2;
+sleep 2000
+
+print =============== create drop qnode 1
+sql create qnode on dnode 1
+sql create snode on dnode 1
+sql create bnode on dnode 1
+
+return
+print =============== restart
+system sh/exec.sh -n dnode1 -s stop -x SIGINT
+system sh/exec.sh -n dnode1 -s start
+
+
+return 
+system sh/deploy.sh -n dnode2 -i 2
+system sh/exec.sh -n dnode2 -s start
+system sh/exec.sh -n dnode2 -s stop -x SIGINT
+system sh/exec.sh -n dnode2 -s start
+
+system sh/exec.sh -n dnode1 -s stop -x SIGINT
+system sh/exec.sh -n dnode2 -s stop -x SIGINT