vmInt.c 11.9 KB
Newer Older
S
shm  
Shengliang Guan 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 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/>.
 */

#define _DEFAULT_SOURCE
S
shm  
Shengliang Guan 已提交
17
#include "vmFile.h"
S
rename  
Shengliang Guan 已提交
18
#include "vmMsg.h"
S
shm  
Shengliang Guan 已提交
19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262
#include "vmWorker.h"

SVnodeObj *vmAcquireVnode(SVnodesMgmt *pMgmt, int32_t vgId) {
  SVnodeObj *pVnode = NULL;
  int32_t    refCount = 0;

  taosRLockLatch(&pMgmt->latch);
  taosHashGetDup(pMgmt->hash, &vgId, sizeof(int32_t), (void *)&pVnode);
  if (pVnode == NULL) {
    terrno = TSDB_CODE_VND_INVALID_VGROUP_ID;
  } else {
    refCount = atomic_add_fetch_32(&pVnode->refCount, 1);
  }
  taosRUnLockLatch(&pMgmt->latch);

  if (pVnode != NULL) {
    dTrace("vgId:%d, acquire vnode, refCount:%d", pVnode->vgId, refCount);
  }

  return pVnode;
}

void vmReleaseVnode(SVnodesMgmt *pMgmt, SVnodeObj *pVnode) {
  if (pVnode == NULL) return;

  SVnodesMgmt *pMgmt = &pDnode->vmgmt;
  taosRLockLatch(&pMgmt->latch);
  int32_t refCount = atomic_sub_fetch_32(&pVnode->refCount, 1);
  taosRUnLockLatch(&pMgmt->latch);
  dTrace("vgId:%d, release vnode, refCount:%d", pVnode->vgId, refCount);
}

int32_t vmOpenVnode(SVnodesMgmt *pMgmt, SWrapperCfg *pCfg, SVnode *pImpl) {
  SVnodeObj *pVnode = calloc(1, sizeof(SVnodeObj));
  if (pVnode == NULL) {
    terrno = TSDB_CODE_OUT_OF_MEMORY;
    return -1;
  }

  pVnode->vgId = pCfg->vgId;
  pVnode->refCount = 0;
  pVnode->dropped = 0;
  pVnode->accessState = TSDB_VN_ALL_ACCCESS;
  pVnode->pImpl = pImpl;
  pVnode->vgVersion = pCfg->vgVersion;
  pVnode->dbUid = pCfg->dbUid;
  pVnode->db = tstrdup(pCfg->db);
  pVnode->path = tstrdup(pCfg->path);

  if (pVnode->path == NULL || pVnode->db == NULL) {
    terrno = TSDB_CODE_OUT_OF_MEMORY;
    return -1;
  }

  if (dndAllocVnodeQueue(pDnode, pVnode) != 0) {
    terrno = TSDB_CODE_OUT_OF_MEMORY;
    return -1;
  }

  taosWLockLatch(&pMgmt->latch);
  int32_t code = taosHashPut(pMgmt->hash, &pVnode->vgId, sizeof(int32_t), &pVnode, sizeof(SVnodeObj *));
  taosWUnLockLatch(&pMgmt->latch);

  if (code != 0) {
    terrno = TSDB_CODE_OUT_OF_MEMORY;
  }
  return code;
}

void vmCloseVnode(SVnodesMgmt *pMgmt, SVnodeObj *pVnode) {
  SVnodesMgmt *pMgmt = &pDnode->vmgmt;
  taosWLockLatch(&pMgmt->latch);
  taosHashRemove(pMgmt->hash, &pVnode->vgId, sizeof(int32_t));
  taosWUnLockLatch(&pMgmt->latch);

  vmReleaseVnode(pMgmt, pVnode);
  while (pVnode->refCount > 0) taosMsleep(10);
  while (!taosQueueEmpty(pVnode->pWriteQ)) taosMsleep(10);
  while (!taosQueueEmpty(pVnode->pSyncQ)) taosMsleep(10);
  while (!taosQueueEmpty(pVnode->pApplyQ)) taosMsleep(10);
  while (!taosQueueEmpty(pVnode->pQueryQ)) taosMsleep(10);
  while (!taosQueueEmpty(pVnode->pFetchQ)) taosMsleep(10);

  dndFreeVnodeQueue(pDnode, pVnode);
  vnodeClose(pVnode->pImpl);
  pVnode->pImpl = NULL;

  dDebug("vgId:%d, vnode is closed", pVnode->vgId);

  if (pVnode->dropped) {
    dDebug("vgId:%d, vnode is destroyed for dropped:%d", pVnode->vgId, pVnode->dropped);
    vnodeDestroy(pVnode->path);
  }

  free(pVnode->path);
  free(pVnode->db);
  free(pVnode);
}

static void *vmOpenVnodeFunc(void *param) {
  SVnodeThread *pThread = param;
  SVnodesMgmt  *pMgmt = pThread->pMgmt;
  SDnode       *pDnode = pMgmt->pDnode;

  dDebug("thread:%d, start to open %d vnodes", pThread->threadIndex, pThread->vnodeNum);
  setThreadName("open-vnodes");

  for (int32_t v = 0; v < pThread->vnodeNum; ++v) {
    SWrapperCfg *pCfg = &pThread->pCfgs[v];

    char stepDesc[TSDB_STEP_DESC_LEN] = {0};
    snprintf(stepDesc, TSDB_STEP_DESC_LEN, "vgId:%d, start to restore, %d of %d have been opened", pCfg->vgId,
             pMgmt->state.openVnodes, pMgmt->state.totalVnodes);
    dndReportStartup(pDnode, "open-vnodes", stepDesc);

    SVnodeCfg cfg = {.pDnode = pDnode, .pTfs = pMgmt->pTfs, .vgId = pCfg->vgId, .dbId = pCfg->dbUid};
    SVnode   *pImpl = vnodeOpen(pCfg->path, &cfg);
    if (pImpl == NULL) {
      dError("vgId:%d, failed to open vnode by thread:%d", pCfg->vgId, pThread->threadIndex);
      pThread->failed++;
    } else {
      vmOpenVnode(pDnode, pCfg, pImpl);
      dDebug("vgId:%d, is opened by thread:%d", pCfg->vgId, pThread->threadIndex);
      pThread->opened++;
    }

    atomic_add_fetch_32(&pMgmt->state.openVnodes, 1);
  }

  dDebug("thread:%d, total vnodes:%d, opened:%d failed:%d", pThread->threadIndex, pThread->vnodeNum, pThread->opened,
         pThread->failed);
  return NULL;
}

static int32_t vmOpenVnodes(SVnodesMgmt *pMgmt) {
  SDnode *pDnode = pMgmt->pDnode;
  taosInitRWLatch(&pMgmt->latch);

  pMgmt->hash = taosHashInit(TSDB_MIN_VNODES, taosGetDefaultHashFunction(TSDB_DATA_TYPE_INT), true, HASH_NO_LOCK);
  if (pMgmt->hash == NULL) {
    dError("failed to init vnode hash");
    terrno = TSDB_CODE_OUT_OF_MEMORY;
    return -1;
  }

  SWrapperCfg *pCfgs = NULL;
  int32_t      numOfVnodes = 0;
  if (vmGetVnodesFromFile(pDnode, &pCfgs, &numOfVnodes) != 0) {
    dInfo("failed to get vnode list from disk since %s", terrstr());
    return -1;
  }

  pMgmt->state.totalVnodes = numOfVnodes;

#if 0 
  int32_t threadNum = tsNumOfCores;
#else
  int32_t threadNum = 1;
#endif
  int32_t vnodesPerThread = numOfVnodes / threadNum + 1;

  SVnodeThread *threads = calloc(threadNum, sizeof(SVnodeThread));
  for (int32_t t = 0; t < threadNum; ++t) {
    threads[t].threadIndex = t;
    threads[t].pMgmt = pMgmt;
    threads[t].pCfgs = calloc(vnodesPerThread, sizeof(SWrapperCfg));
  }

  for (int32_t v = 0; v < numOfVnodes; ++v) {
    int32_t       t = v % threadNum;
    SVnodeThread *pThread = &threads[t];
    pThread->pCfgs[pThread->vnodeNum++] = pCfgs[v];
  }

  dInfo("start %d threads to open %d vnodes", threadNum, numOfVnodes);

  for (int32_t t = 0; t < threadNum; ++t) {
    SVnodeThread *pThread = &threads[t];
    if (pThread->vnodeNum == 0) continue;

    pthread_attr_t thAttr;
    pthread_attr_init(&thAttr);
    pthread_attr_setdetachstate(&thAttr, PTHREAD_CREATE_JOINABLE);
    if (pthread_create(&pThread->thread, &thAttr, vmOpenVnodeFunc, pThread) != 0) {
      dError("thread:%d, failed to create thread to open vnode, reason:%s", pThread->threadIndex, strerror(errno));
    }

    pthread_attr_destroy(&thAttr);
  }

  for (int32_t t = 0; t < threadNum; ++t) {
    SVnodeThread *pThread = &threads[t];
    if (pThread->vnodeNum > 0 && taosCheckPthreadValid(pThread->thread)) {
      pthread_join(pThread->thread, NULL);
    }
    free(pThread->pCfgs);
  }
  free(threads);
  free(pCfgs);

  if (pMgmt->state.openVnodes != pMgmt->state.totalVnodes) {
    dError("there are total vnodes:%d, opened:%d", pMgmt->state.totalVnodes, pMgmt->state.openVnodes);
    return -1;
  } else {
    dInfo("total vnodes:%d open successfully", pMgmt->state.totalVnodes);
    return 0;
  }
}

static void vmCloseVnodes(SVnodesMgmt *pMgmt) {
  dInfo("start to close all vnodes");

  int32_t     numOfVnodes = 0;
  SVnodeObj **pVnodes = dndGetVnodesFromHash(pDnode, &numOfVnodes);

  for (int32_t i = 0; i < numOfVnodes; ++i) {
    vmCloseVnode(pDnode, pVnodes[i]);
  }

  if (pVnodes != NULL) {
    free(pVnodes);
  }

  if (pMgmt->hash != NULL) {
    taosHashCleanup(pMgmt->hash);
    pMgmt->hash = NULL;
  }

  dInfo("total vnodes:%d are all closed", numOfVnodes);
}

static void vmCleanup(SMgmtWrapper *pWrapper) {
  SVnodesMgmt *pMgmt = pWrapper->pMgmt;
  if (pMgmt == NULL) return;

  dInfo("vnodes-mgmt start to cleanup");
  vmCloseVnodes(pMgmt);
  vmStopWorker(pMgmt);
  vnodeCleanup();
  // walCleanUp();
  free(pMgmt);
  pWrapper->pMgmt = NULL;
  dInfo("vnodes-mgmt is cleaned up");
}
S
shm  
Shengliang Guan 已提交
263

S
shm  
Shengliang Guan 已提交
264
static int32_t vmInit(SMgmtWrapper *pWrapper) {
S
shm  
Shengliang Guan 已提交
265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294
  SDnode      *pDnode = pWrapper->pDnode;
  SVnodesMgmt *pMgmt = calloc(1, sizeof(SVnodesMgmt));
  int32_t      code = -1;
  SVnodeOpt    vnodeOpt = {0};

  dInfo("vnodes-mgmt start to init");
  if (pMgmt == NULL) goto _OVER;

  pMgmt->path = pWrapper->path;
  pMgmt->pDnode = pWrapper->pDnode;
  pMgmt->pWrapper = pWrapper;
  taosInitRWLatch(&pMgmt->latch);

  SDiskCfg dCfg = {0};
  tstrncpy(dCfg.dir, pDnode->cfg.dataDir, TSDB_FILENAME_LEN);
  dCfg.level = 0;
  dCfg.primary = 1;
  SDiskCfg *pDisks = pDnode->cfg.pDisks;
  int32_t   numOfDisks = pDnode->cfg.numOfDisks;
  if (numOfDisks <= 0 || pDisks == NULL) {
    pDisks = &dCfg;
    numOfDisks = 1;
  }

  pMgmt->pTfs = tfsOpen(pDisks, numOfDisks);
  if (pMgmt->pTfs == NULL) {
    dError("failed to init tfs since %s", terrstr());
    goto _OVER;
  }

S
shm  
Shengliang Guan 已提交
295 296
  if (walInit() != 0) {
    dError("failed to init wal since %s", terrstr());
S
shm  
Shengliang Guan 已提交
297
    goto _OVER;
S
shm  
Shengliang Guan 已提交
298
  }
S
shm  
Shengliang Guan 已提交
299

S
shm  
Shengliang Guan 已提交
300 301 302 303 304
  vnodeOpt.nthreads = tsNumOfCommitThreads;
  vnodeOpt.putReqToVQueryQFp = dndPutReqToVQueryQ;
  vnodeOpt.sendReqToDnodeFp = dndSendReqToDnode;
  if (vnodeInit(&vnodeOpt) != 0) {
    dError("failed to init vnode since %s", terrstr());
S
shm  
Shengliang Guan 已提交
305 306 307 308 309 310 311 312 313
    goto _OVER;
  }

  if (vmStartWorker(pMgmt) != 0) {
    dError("failed to init workers since %s", terrstr()) goto _OVER;
  }

  if (vmOpenVnodes(pMgmt) != 0) {
    dError("failed to open vnodes since %s", terrstr());
S
shm  
Shengliang Guan 已提交
314 315
    return -1;
  }
S
shm  
Shengliang Guan 已提交
316

S
shm  
Shengliang Guan 已提交
317 318 319 320 321 322 323 324
_OVER:
  if (code == 0) {
    pWrapper->pMgmt = pMgmt;
    dInfo("vnodes-mgmt is initialized");
  } else {
    dError("failed to init vnodes-mgmt since %s", terrstr());
    vmCleanup(pWrapper);
  }
S
shm  
Shengliang Guan 已提交
325

S
shm  
Shengliang Guan 已提交
326
  return 0;
S
shm  
Shengliang Guan 已提交
327 328 329
}

static bool vmRequire(SMgmtWrapper *pWrapper) { return false; }
S
shm  
Shengliang Guan 已提交
330

S
shm  
Shengliang Guan 已提交
331
void vmGetMgmtFp(SMgmtWrapper *pWrapper) {
S
shm  
Shengliang Guan 已提交
332
  SMgmtFp mgmtFp = {0};
S
shm  
Shengliang Guan 已提交
333 334 335
  mgmtFp.openFp = vmInit;
  mgmtFp.closeFp = vmCleanup;
  mgmtFp.requiredFp = vmRequire;
S
shm  
Shengliang Guan 已提交
336 337 338 339

  vmInitMsgHandles(pWrapper);
  pWrapper->name = "vnodes";
  pWrapper->fp = mgmtFp;
S
shm  
Shengliang Guan 已提交
340
}
S
shm  
Shengliang Guan 已提交
341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408

int32_t vmGetTfsMonitorInfo(SMgmtWrapper *pWrapper, SMonDiskInfo *pInfo) {
  SVnodesMgmt *pMgmt = pWrapper->pMgmt;
  if (pMgmt == NULL) return -1;

  return tfsGetMonitorInfo(pMgmt->pTfs, pInfo);
  ;
}

void vmGetVndMonitorInfo(SMgmtWrapper *pWrapper, SMonDnodeInfo *pInfo) {
  SVnodesMgmt *pMgmt = pWrapper->pMgmt;
  if (pMgmt == NULL) return;

  SVnodesStat *pStat = &pMgmt->state;
  pInfo->req_select = pStat->numOfSelectReqs;
  pInfo->req_insert = pStat->numOfInsertReqs;
  pInfo->req_insert_success = pStat->numOfInsertSuccessReqs;
  pInfo->req_insert_batch = pStat->numOfBatchInsertReqs;
  pInfo->req_insert_batch_success = pStat->numOfBatchInsertSuccessReqs;
  pInfo->errors = tsNumOfErrorLogs;
  pInfo->vnodes_num = pStat->totalVnodes;
  pInfo->masters = pStat->masterNum;
}

void vmGetVnodeLoads(SDnode *pDnode, SArray *pLoads) {
  SVnodesMgmt *pMgmt = &pDnode->vmgmt;
  SVnodesStat *pStat = &pMgmt->stat;
  int32_t      totalVnodes = 0;
  int32_t      masterNum = 0;
  int64_t      numOfSelectReqs = 0;
  int64_t      numOfInsertReqs = 0;
  int64_t      numOfInsertSuccessReqs = 0;
  int64_t      numOfBatchInsertReqs = 0;
  int64_t      numOfBatchInsertSuccessReqs = 0;

  taosRLockLatch(&pMgmt->latch);

  void *pIter = taosHashIterate(pMgmt->hash, NULL);
  while (pIter) {
    SVnodeObj **ppVnode = pIter;
    if (ppVnode == NULL || *ppVnode == NULL) continue;

    SVnodeObj *pVnode = *ppVnode;
    SVnodeLoad vload = {0};
    vnodeGetLoad(pVnode->pImpl, &vload);
    taosArrayPush(pLoads, &vload);

    numOfSelectReqs += vload.numOfSelectReqs;
    numOfInsertReqs += vload.numOfInsertReqs;
    numOfInsertSuccessReqs += vload.numOfInsertSuccessReqs;
    numOfBatchInsertReqs += vload.numOfBatchInsertReqs;
    numOfBatchInsertSuccessReqs += vload.numOfBatchInsertSuccessReqs;
    totalVnodes++;
    if (vload.role == TAOS_SYNC_STATE_LEADER) masterNum++;

    pIter = taosHashIterate(pMgmt->hash, pIter);
  }

  taosRUnLockLatch(&pMgmt->latch);

  pStat->totalVnodes = totalVnodes;
  pStat->masterNum = masterNum;
  pStat->numOfSelectReqs = numOfSelectReqs;
  pStat->numOfInsertReqs = numOfInsertReqs;
  pStat->numOfInsertSuccessReqs = numOfInsertSuccessReqs;
  pStat->numOfBatchInsertReqs = numOfBatchInsertReqs;
  pStat->numOfBatchInsertSuccessReqs = numOfBatchInsertSuccessReqs;
}