crash_gen.py 54.1 KB
Newer Older
1
#!/usr/bin/python3.7
S
Steven Li 已提交
2 3 4 5 6 7 8 9 10 11 12 13
###################################################################
#           Copyright (c) 2016 by TAOS Technologies, Inc.
#                     All rights reserved.
#
#  This file is proprietary and confidential to TAOS Technologies.
#  No part of this file may be reproduced, stored, transmitted,
#  disclosed or used in any form or by any means other than as
#  expressly provided by the written permission from Jianhui Tao
#
###################################################################

# -*- coding: utf-8 -*-
14 15
from __future__ import annotations  # For type hinting before definition, ref: https://stackoverflow.com/questions/33533148/how-do-i-specify-that-the-return-type-of-a-method-is-the-same-as-the-class-itsel    

S
Steven Li 已提交
16
import sys
17
import traceback
18 19 20 21
# Require Python 3
if sys.version_info[0] < 3:
    raise Exception("Must be using Python 3")

S
Steven Li 已提交
22
import getopt
23
import argparse
24
import copy
S
Steven Li 已提交
25 26 27

import threading
import random
28
import time
S
Steven Li 已提交
29
import logging
30
import datetime
31
import textwrap
S
Steven Li 已提交
32

33
from typing import List
34
from typing import Dict
35

S
Steven Li 已提交
36 37 38 39 40
from util.log import *
from util.dnodes import *
from util.cases import *
from util.sql import *

41
import crash_gen
S
Steven Li 已提交
42 43
import taos

44 45 46
# Global variables, tried to keep a small number. 
gConfig = None # Command-line/Environment Configurations, will set a bit later
logger = None
S
Steven Li 已提交
47

48 49
def runThread(wt: WorkerThread):    
    wt.run()
50

51 52 53 54 55 56 57 58
class CrashGenError(Exception):
    def __init__(self, msg=None, errno=None):
        self.msg = msg    
        self.errno = errno
    
    def __str__(self):
        return self.msg

S
Steven Li 已提交
59
class WorkerThread:
60
    def __init__(self, pool: ThreadPool, tid, 
61 62 63 64
            tc: ThreadCoordinator,
            # te: TaskExecutor,
            ): # note: main thread context!
        # self._curStep = -1 
65
        self._pool = pool
66
        self._tid = tid        
67
        self._tc = tc
S
Steven Li 已提交
68
        # self.threadIdent = threading.get_ident()
69 70
        self._thread = threading.Thread(target=runThread, args=(self,))
        self._stepGate = threading.Event()
S
Steven Li 已提交
71

72
        # Let us have a DB connection of our own
73 74 75
        if ( gConfig.per_thread_db_connection ): # type: ignore
            self._dbConn = DbConn()   

76
    def logDebug(self, msg):
S
Steven Li 已提交
77
        logger.debug("    TRD[{}] {}".format(self._tid, msg))
78 79

    def logInfo(self, msg):
S
Steven Li 已提交
80
        logger.info("    TRD[{}] {}".format(self._tid, msg))
81 82

   
83 84
    def getTaskExecutor(self):
        return self._tc.getTaskExecutor()     
85

S
Steven Li 已提交
86
    def start(self):
87
        self._thread.start()  # AFTER the thread is recorded
S
Steven Li 已提交
88

89
    def run(self): 
S
Steven Li 已提交
90
        # initialization after thread starts, in the thread context
91
        # self.isSleeping = False
92 93
        logger.info("Starting to run thread: {}".format(self._tid))

94
        if ( gConfig.per_thread_db_connection ): # type: ignore
95
            self._dbConn.open()
S
Steven Li 已提交
96

97 98
        self._doTaskLoop()       
        
99
        # clean up
100
        if ( gConfig.per_thread_db_connection ): # type: ignore 
101
            self._dbConn.close()
102

103 104 105
    def _doTaskLoop(self) :
        # while self._curStep < self._pool.maxSteps:
        # tc = ThreadCoordinator(None)
106 107 108
        while True:  
            tc = self._tc # Thread Coordinator, the overall master            
            tc.crossStepBarrier()  # shared barrier first, INCLUDING the last one
S
Steven Li 已提交
109
            logger.debug("[TRD] Worker thread [{}] exited barrier...".format(self._tid))
110
            self.crossStepGate()   # then per-thread gate, after being tapped
S
Steven Li 已提交
111
            logger.debug("[TRD] Worker thread [{}] exited step gate...".format(self._tid))
112
            if not self._tc.isRunning():
S
Steven Li 已提交
113
                logger.debug("[TRD] Thread Coordinator not running any more, worker thread now stopping...")
114 115
                break

S
Steven Li 已提交
116
            logger.debug("[TRD] Worker thread [{}] about to fetch task".format(self._tid))
117
            task = tc.fetchTask()
S
Steven Li 已提交
118
            logger.debug("[TRD] Worker thread [{}] about to execute task: {}".format(self._tid, task.__class__.__name__))
119
            task.execute(self)
120
            tc.saveExecutedTask(task)
S
Steven Li 已提交
121
            logger.debug("[TRD] Worker thread [{}] finished executing task".format(self._tid))
122
  
S
Steven Li 已提交
123
    def verifyThreadSelf(self): # ensure we are called by this own thread
124
        if ( threading.get_ident() != self._thread.ident ): 
S
Steven Li 已提交
125 126 127 128 129 130 131
            raise RuntimeError("Unexpectly called from other threads")

    def verifyThreadMain(self): # ensure we are called by the main thread
        if ( threading.get_ident() != threading.main_thread().ident ): 
            raise RuntimeError("Unexpectly called from other threads")

    def verifyThreadAlive(self):
132
        if ( not self._thread.is_alive() ):
S
Steven Li 已提交
133 134
            raise RuntimeError("Unexpected dead thread")

135
    # A gate is different from a barrier in that a thread needs to be "tapped"
S
Steven Li 已提交
136 137 138 139
    def crossStepGate(self):
        self.verifyThreadAlive()
        self.verifyThreadSelf() # only allowed by ourselves
        
140
        # Wait again at the "gate", waiting to be "tapped"
S
Steven Li 已提交
141
        logger.debug("[TRD] Worker thread {} about to cross the step gate".format(self._tid))
142 143
        self._stepGate.wait() 
        self._stepGate.clear()
S
Steven Li 已提交
144
        
145
        # self._curStep += 1  # off to a new step...
S
Steven Li 已提交
146 147 148 149

    def tapStepGate(self): # give it a tap, release the thread waiting there
        self.verifyThreadAlive()
        self.verifyThreadMain() # only allowed for main thread
150
 
S
Steven Li 已提交
151
        logger.debug("[TRD] Tapping worker thread {}".format(self._tid))
152 153
        self._stepGate.set() # wake up!        
        time.sleep(0) # let the released thread run a bit
154

155
    def execSql(self, sql): # TODO: expose DbConn directly
156 157 158
        if ( gConfig.per_thread_db_connection ):
            return self._dbConn.execute(sql)            
        else:
159
            return self._tc.getDbState().getDbConn().execute(sql)
160

161 162 163 164 165 166
    def getDbConn(self):
        if ( gConfig.per_thread_db_connection ):
            return self._dbConn     
        else:
            return self._tc.getDbState().getDbConn()

167 168 169 170 171
    # def querySql(self, sql): # not "execute", since we are out side the DB context
    #     if ( gConfig.per_thread_db_connection ):
    #         return self._dbConn.query(sql)            
    #     else:
    #         return self._tc.getDbState().getDbConn().query(sql)
172

173
class ThreadCoordinator:
174
    def __init__(self, pool, dbState):
175 176
        self._curStep = -1 # first step is 0
        self._pool = pool
177
        # self._wd = wd
178
        self._te = None # prepare for every new step
179 180 181
        self._dbState = dbState
        self._executedTasks: List[Task] = [] # in a given step
        self._lock = threading.RLock() # sync access for a few things
S
Steven Li 已提交
182

183
        self._stepBarrier = threading.Barrier(self._pool.numThreads + 1) # one barrier for all threads
184
        self._execStats = ExecutionStats()
S
Steven Li 已提交
185

186 187 188
    def getTaskExecutor(self):
        return self._te

189 190 191
    def getDbState(self) -> DbState :
        return self._dbState

192 193 194
    def crossStepBarrier(self):
        self._stepBarrier.wait()

195 196
    def run(self):              
        self._pool.createAndStartThreads(self)
S
Steven Li 已提交
197 198

        # Coordinate all threads step by step
199 200
        self._curStep = -1 # not started yet
        maxSteps = gConfig.max_steps # type: ignore
201 202 203
        self._execStats.startExec() # start the stop watch
        failed = False
        while(self._curStep < maxSteps-1 and not failed):  # maxStep==10, last curStep should be 9
S
Steven Li 已提交
204 205 206
            if not gConfig.debug: 
                print(".", end="", flush=True) # print this only if we are not in debug mode
            logger.debug("[TRD] Main thread going to sleep")
207

208
            # Now ready to enter a step
209 210 211 212
            self.crossStepBarrier() # let other threads go past the pool barrier, but wait at the thread gate
            self._stepBarrier.reset() # Other worker threads should now be at the "gate"            

            # At this point, all threads should be pass the overall "barrier" and before the per-thread "gate"
213 214 215 216 217 218 219 220 221 222 223 224 225 226 227
            try:
                self._dbState.transition(self._executedTasks) # at end of step, transiton the DB state
            except taos.error.ProgrammingError as err:
                if ( err.msg == 'network unavailable' ): # broken DB connection
                    logger.info("DB connection broken, execution failed")
                    traceback.print_stack()
                    failed = True
                    self._te = None # Not running any more
                    self._execStats.registerFailure("Broken DB Connection")
                    # continue # don't do that, need to tap all threads at end, and maybe signal them to stop
                else:
                    raise 
            finally:
                pass
            
228
            self.resetExecutedTasks() # clear the tasks after we are done
229 230

            # Get ready for next step
S
Steven Li 已提交
231
            logger.debug("<-- Step {} finished".format(self._curStep))
232 233
            self._curStep += 1 # we are about to get into next step. TODO: race condition here!                
            logger.debug("\r\n--> Step {} starts with main thread waking up".format(self._curStep)) # Now not all threads had time to go to sleep
234

235
            # A new TE for the new step
236 237
            if not failed: # only if not failed
                self._te = TaskExecutor(self._curStep)
238

S
Steven Li 已提交
239
            logger.debug("[TRD] Main thread waking up at step {}, tapping worker threads".format(self._curStep)) # Now not all threads had time to go to sleep            
S
Steven Li 已提交
240 241
            self.tapAllThreads()

242
        logger.debug("Main thread ready to finish up...")
243 244 245 246 247 248 249 250
        if not failed: # only in regular situations
            self.crossStepBarrier() # Cross it one last time, after all threads finish
            self._stepBarrier.reset()
            logger.debug("Main thread in exclusive zone...")
            self._te = None # No more executor, time to end
            logger.debug("Main thread tapping all threads one last time...")
            self.tapAllThreads() # Let the threads run one last time

251 252
        logger.debug("Main thread joining all threads")
        self._pool.joinAll() # Get all threads to finish
S
Steven Li 已提交
253
        logger.info("All worker thread finished")
254 255 256
        self._execStats.endExec()

    def logStats(self):
257
        self._execStats.logStats()
S
Steven Li 已提交
258 259 260

    def tapAllThreads(self): # in a deterministic manner
        wakeSeq = []
261
        for i in range(self._pool.numThreads): # generate a random sequence
S
Steven Li 已提交
262 263 264 265
            if Dice.throw(2) == 1 :
                wakeSeq.append(i)
            else:
                wakeSeq.insert(0, i)
S
Steven Li 已提交
266
        logger.debug("[TRD] Main thread waking up worker thread: {}".format(str(wakeSeq)))
267
        # TODO: set dice seed to a deterministic value
S
Steven Li 已提交
268
        for i in wakeSeq:
269
            self._pool.threadList[i].tapStepGate() # TODO: maybe a bit too deep?!
S
Steven Li 已提交
270 271
            time.sleep(0) # yield

272 273 274 275 276 277
    def isRunning(self):
        return self._te != None

    def fetchTask(self) -> Task :
        if ( not self.isRunning() ): # no task
            raise RuntimeError("Cannot fetch task when not running")
278 279
        # return self._wd.pickTask()
        # Alternatively, let's ask the DbState for the appropriate task
280 281 282 283 284 285 286 287 288
        # dbState = self.getDbState()
        # tasks = dbState.getTasksAtState() # TODO: create every time?
        # nTasks = len(tasks)
        # i = Dice.throw(nTasks)
        # logger.debug(" (dice:{}/{}) ".format(i, nTasks))
        # # return copy.copy(tasks[i]) # Needs a fresh copy, to save execution results, etc.
        # return tasks[i].clone() # TODO: still necessary?
        taskType = self.getDbState().pickTaskType() # pick a task type for current state
        return taskType(self.getDbState(), self._execStats) # create a task from it
289 290 291

    def resetExecutedTasks(self):
        self._executedTasks = [] # should be under single thread
292 293 294 295

    def saveExecutedTask(self, task):
        with self._lock:
            self._executedTasks.append(task)
296 297

# We define a class to run a number of threads in locking steps.
298
class ThreadPool:
299 300 301 302 303
    def __init__(self, dbState, numThreads, maxSteps, funcSequencer):
        self.numThreads = numThreads
        self.maxSteps = maxSteps
        self.funcSequencer = funcSequencer
        # Internal class variables
304
        # self.dispatcher = WorkDispatcher(dbState) # Obsolete?
305 306 307
        self.curStep = 0
        self.threadList = []
        # self.stepGate = threading.Condition() # Gate to hold/sync all threads
308
        # self.numWaitingThreads = 0    
309 310
        
    # starting to run all the threads, in locking steps
311
    def createAndStartThreads(self, tc: ThreadCoordinator):
312
        for tid in range(0, self.numThreads): # Create the threads
313
            workerThread = WorkerThread(self, tid, tc)            
314 315 316 317 318 319 320 321
            self.threadList.append(workerThread)
            workerThread.start() # start, but should block immediately before step 0

    def joinAll(self):
        for workerThread in self.threadList:
            logger.debug("Joining thread...")
            workerThread._thread.join()

S
Steven Li 已提交
322 323 324
# A queue of continguous POSITIVE integers
class LinearQueue():
    def __init__(self):
325
        self.firstIndex = 1  # 1st ever element
S
Steven Li 已提交
326
        self.lastIndex = 0
327
        self._lock = threading.RLock() # our functions may call each other
328
        self.inUse = set() # the indexes that are in use right now
S
Steven Li 已提交
329

330 331 332 333 334 335 336 337 338
    def toText(self):
        return "[{}..{}], in use: {}".format(self.firstIndex, self.lastIndex, self.inUse)

    # Push (add new element, largest) to the tail, and mark it in use
    def push(self): 
        with self._lock:
            # if ( self.isEmpty() ): 
            #     self.lastIndex = self.firstIndex 
            #     return self.firstIndex
339 340
            # Otherwise we have something
            self.lastIndex += 1
341 342
            self.allocate(self.lastIndex)
            # self.inUse.add(self.lastIndex) # mark it in use immediately
343
            return self.lastIndex
S
Steven Li 已提交
344 345

    def pop(self):
346
        with self._lock:
347
            if ( self.isEmpty() ): 
348 349 350
                # raise RuntimeError("Cannot pop an empty queue") 
                return False # TODO: None?
            
351
            index = self.firstIndex
352
            if ( index in self.inUse ):
353 354
                return False

355 356 357 358 359 360 361
            self.firstIndex += 1
            return index

    def isEmpty(self):
        return self.firstIndex > self.lastIndex

    def popIfNotEmpty(self):
362
        with self._lock:
363 364 365 366
            if (self.isEmpty()):
                return 0
            return self.pop()

S
Steven Li 已提交
367
    def allocate(self, i):
368
        with self._lock:
369
            # logger.debug("LQ allocating item {}".format(i))
370 371 372 373
            if ( i in self.inUse ):
                raise RuntimeError("Cannot re-use same index in queue: {}".format(i))
            self.inUse.add(i)

S
Steven Li 已提交
374
    def release(self, i):
375
        with self._lock:
376 377
            # logger.debug("LQ releasing item {}".format(i))
            self.inUse.remove(i) # KeyError possible, TODO: why?
378 379 380 381

    def size(self):
        return self.lastIndex + 1 - self.firstIndex

S
Steven Li 已提交
382
    def pickAndAllocate(self):
383 384 385
        if ( self.isEmpty() ):
            return None
        with self._lock:
386 387 388 389
            cnt = 0 # counting the interations
            while True:
                cnt += 1
                if ( cnt > self.size()*10 ): # 10x iteration already
390 391
                    # raise RuntimeError("Failed to allocate LinearQueue element")
                    return None
392 393
                ret = Dice.throwRange(self.firstIndex, self.lastIndex+1)
                if ( not ret in self.inUse ):
394 395 396 397 398
                    self.allocate(ret)
                    return ret

class DbConn:
    def __init__(self):
399 400
        self._conn = None 
        self._cursor = None
401 402 403 404 405 406 407
        self.isOpen = False
        
    def open(self): # Open connection
        if ( self.isOpen ):
            raise RuntimeError("Cannot re-open an existing DB connection")

        cfgPath = "../../build/test/cfg" 
408 409
        self._conn = taos.connect(host="127.0.0.1", config=cfgPath) # TODO: make configurable
        self._cursor = self._conn.cursor()
410

411 412
        # Get the connection/cursor ready
        self._cursor.execute('reset query cache')
413
        # self._cursor.execute('use db') # note we do this in _findCurrenState
414 415

        # Open connection
416
        self._tdSql = TDSql()
417
        self._tdSql.init(self._cursor)
418 419 420 421 422
        self.isOpen = True

    def resetDb(self): # reset the whole database, etc.
        if ( not self.isOpen ):
            raise RuntimeError("Cannot reset database until connection is open")
423 424 425
        # self._tdSql.prepare() # Recreate database, etc.

        self._cursor.execute('drop database if exists db')
426 427
        logger.debug("Resetting DB, dropped database")
        # self._cursor.execute('create database db')
428 429
        # self._cursor.execute('use db')

430 431 432 433 434 435 436
        # tdSql.execute('show databases')

    def close(self):
        if ( not self.isOpen ):
            raise RuntimeError("Cannot clean up database until connection is open")
        self._tdSql.close()
        self.isOpen = False
S
Steven Li 已提交
437

438
    def execute(self, sql): 
439
        if ( not self.isOpen ):
440
            raise RuntimeError("Cannot execute database commands until connection is open")
441 442 443 444
        logger.debug("[SQL] Executing SQL: {}".format(sql))
        nRows = self._tdSql.execute(sql)
        logger.debug("[SQL] Execution Result, nRows = {}, SQL = {}".format(nRows, sql))
        return nRows
S
Steven Li 已提交
445

446
    def query(self, sql) :  # return rows affected
447 448
        if ( not self.isOpen ):
            raise RuntimeError("Cannot query database until connection is open")
449 450 451 452
        logger.debug("[SQL] Executing SQL: {}".format(sql))
        nRows = self._tdSql.query(sql)
        logger.debug("[SQL] Execution Result, nRows = {}, SQL = {}".format(nRows, sql))
        return nRows
453
        # results are in: return self._tdSql.queryResult
454

455 456 457
    def getQueryResult(self):
        return self._tdSql.queryResult

458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475
    def _queryAny(self, sql) : # actual query result as an int
        if ( not self.isOpen ):
            raise RuntimeError("Cannot query database until connection is open")
        tSql = self._tdSql
        nRows = tSql.query(sql)
        if nRows != 1 :
            raise RuntimeError("Unexpected result for query: {}, rows = {}".format(sql, nRows))
        if tSql.queryRows != 1 or tSql.queryCols != 1:
            raise RuntimeError("Unexpected result set for query: {}".format(sql))
        return tSql.queryResult[0][0]

    def queryScalar(self, sql) -> int :
        return self._queryAny(sql)

    def queryString(self, sql) -> str :
        return self._queryAny(sql)
    
class AnyState:
476
    STATE_INVALID    = -1
477 478 479 480
    STATE_EMPTY      = 0  # nothing there, no even a DB
    STATE_DB_ONLY    = 1  # we have a DB, but nothing else
    STATE_TABLE_ONLY = 2  # we have a table, but totally empty
    STATE_HAS_DATA   = 3  # we have some data in the table
481 482 483 484 485
    _stateNames = ["Invalid", "Empty", "DB_Only", "Table_Only", "Has_Data"]

    STATE_VAL_IDX = 0
    CAN_CREATE_DB = 1
    CAN_DROP_DB = 2
486 487
    CAN_CREATE_FIXED_SUPER_TABLE = 3
    CAN_DROP_FIXED_SUPER_TABLE = 4
488 489 490 491 492 493 494
    CAN_ADD_DATA = 5
    CAN_READ_DATA = 6

    def __init__(self):
        self._info = self.getInfo()

    def __str__(self):
S
Steven Li 已提交
495
        return self._stateNames[self._info[self.STATE_VAL_IDX] + 1] # -1 hack to accomodate the STATE_INVALID case
496 497 498 499

    def getInfo(self):
        raise RuntimeError("Must be overriden by child classes")

S
Steven Li 已提交
500 501 502 503 504 505 506 507
    def equals(self, other):
        if isinstance(other, int):
            return self.getValIndex() == other
        elif isinstance(other, AnyState):
            return self.getValIndex() == other.getValIndex()
        else:
            raise RuntimeError("Unexpected comparison, type = {}".format(type(other)))

508 509 510
    def verifyTasksToState(self, tasks, newState):
        raise RuntimeError("Must be overriden by child classes")

S
Steven Li 已提交
511 512 513
    def getValIndex(self):
        return self._info[self.STATE_VAL_IDX]

514 515 516 517 518 519
    def getValue(self):
        return self._info[self.STATE_VAL_IDX]
    def canCreateDb(self):
        return self._info[self.CAN_CREATE_DB]
    def canDropDb(self):
        return self._info[self.CAN_DROP_DB]
520 521 522 523
    def canCreateFixedSuperTable(self):
        return self._info[self.CAN_CREATE_FIXED_SUPER_TABLE]
    def canDropFixedSuperTable(self):
        return self._info[self.CAN_DROP_FIXED_SUPER_TABLE]
524 525 526 527 528 529 530 531 532 533 534
    def canAddData(self):
        return self._info[self.CAN_ADD_DATA]
    def canReadData(self):
        return self._info[self.CAN_READ_DATA]

    def assertAtMostOneSuccess(self, tasks, cls):
        sCnt = 0
        for task in tasks :
            if not isinstance(task, cls):
                continue
            if task.isSuccess():
S
Steven Li 已提交
535
                # task.logDebug("Task success found")
536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570
                sCnt += 1
                if ( sCnt >= 2 ):
                    raise RuntimeError("Unexpected more than 1 success with task: {}".format(cls))

    def assertIfExistThenSuccess(self, tasks, cls):
        sCnt = 0
        exists = False
        for task in tasks :
            if not isinstance(task, cls):
                continue
            exists = True # we have a valid instance
            if task.isSuccess():
                sCnt += 1
        if ( exists and sCnt <= 0 ):
            raise RuntimeError("Unexpected zero success for task: {}".format(cls))

    def assertNoTask(self, tasks, cls):
        for task in tasks :
            if isinstance(task, cls):
                raise CrashGenError("This task: {}, is not expected to be present, given the success/failure of others".format(cls.__name__))

    def assertNoSuccess(self, tasks, cls):
        for task in tasks :
            if isinstance(task, cls):
                if task.isSuccess():
                    raise RuntimeError("Unexpected successful task: {}".format(cls))

    def hasSuccess(self, tasks, cls):
        for task in tasks :
            if not isinstance(task, cls):
                continue
            if task.isSuccess():
                return True
        return False

S
Steven Li 已提交
571 572 573 574 575 576
    def hasTask(self, tasks, cls):
        for task in tasks :
            if isinstance(task, cls):
                return True
        return False

577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596
class StateInvalid(AnyState):
    def getInfo(self):
        return [
            self.STATE_INVALID,
            False, False, # can create/drop Db
            False, False, # can create/drop fixed table
            False, False, # can insert/read data with fixed table
        ]

    # def verifyTasksToState(self, tasks, newState):

class StateEmpty(AnyState):
    def getInfo(self):
        return [
            self.STATE_EMPTY,
            True, False, # can create/drop Db
            False, False, # can create/drop fixed table
            False, False, # can insert/read data with fixed table
        ]

S
Steven Li 已提交
597 598 599 600
    def verifyTasksToState(self, tasks, newState): 
        if ( self.hasSuccess(tasks, CreateDbTask) ): # at EMPTY, if there's succes in creating DB
            if ( not self.hasTask(tasks, DropDbTask) ) : # and no drop_db tasks
                self.assertAtMostOneSuccess(tasks, CreateDbTask) # we must have at most one. TODO: compare numbers
601 602 603 604 605 606 607 608 609 610 611

class StateDbOnly(AnyState):
    def getInfo(self):
        return [
            self.STATE_DB_ONLY,
            False, True,
            True, False,
            False, False,
        ]

    def verifyTasksToState(self, tasks, newState):
612 613
        if ( not self.hasTask(tasks, CreateDbTask) ):
            self.assertAtMostOneSuccess(tasks, DropDbTask) # only if we don't create any more
614
        self.assertIfExistThenSuccess(tasks, DropDbTask)
S
Steven Li 已提交
615
        # self.assertAtMostOneSuccess(tasks, CreateFixedTableTask) # not true in massively parrallel cases
616 617 618 619 620
        # Nothing to be said about adding data task
        if ( self.hasSuccess(tasks, DropDbTask) ): # dropped the DB
            # self.assertHasTask(tasks, DropDbTask) # implied by hasSuccess
            self.assertAtMostOneSuccess(tasks, DropDbTask)
            # self._state = self.STATE_EMPTY
621
        elif ( self.hasSuccess(tasks, CreateFixedSuperTableTask) ): # did not drop db, create table success
622
            # self.assertHasTask(tasks, CreateFixedTableTask) # tried to create table
623 624
            if ( not self.hasTask(tasks, DropFixedSuperTableTask) ): 
                self.assertAtMostOneSuccess(tasks, CreateFixedSuperTableTask) # at most 1 attempt is successful, if we don't drop anything
625 626 627 628 629 630 631 632 633 634 635 636 637
            self.assertNoTask(tasks, DropDbTask) # should have have tried
            # if ( not self.hasSuccess(tasks, AddFixedDataTask) ): # just created table, no data yet
            #     # can't say there's add-data attempts, since they may all fail
            #     self._state = self.STATE_TABLE_ONLY
            # else:                    
            #     self._state = self.STATE_HAS_DATA
        # What about AddFixedData?
        # elif ( self.hasSuccess(tasks, AddFixedDataTask) ):
        #     self._state = self.STATE_HAS_DATA
        # else: # no success in dropping db tasks, no success in create fixed table? read data should also fail
        #     # raise RuntimeError("Unexpected no-success scenario")   # We might just landed all failure tasks, 
        #     self._state = self.STATE_DB_ONLY  # no change

638
class StateSuperTableOnly(AnyState):
639 640 641 642 643 644 645 646 647
    def getInfo(self):
        return [
            self.STATE_TABLE_ONLY,
            False, True,
            False, True,
            True, True,
        ]

    def verifyTasksToState(self, tasks, newState):
648 649
        if ( self.hasSuccess(tasks, DropFixedSuperTableTask) ): # we are able to drop the table
            self.assertAtMostOneSuccess(tasks, DropFixedSuperTableTask)
650
            # self._state = self.STATE_DB_ONLY
S
Steven Li 已提交
651 652
        # elif ( self.hasSuccess(tasks, AddFixedDataTask) ): # no success dropping the table, but added data
        #     self.assertNoTask(tasks, DropFixedTableTask) # not true in massively parrallel cases
653
            # self._state = self.STATE_HAS_DATA
S
Steven Li 已提交
654 655 656
        # elif ( self.hasSuccess(tasks, ReadFixedDataTask) ): # no success in prev cases, but was able to read data
            # self.assertNoTask(tasks, DropFixedTableTask)
            # self.assertNoTask(tasks, AddFixedDataTask)
657
            # self._state = self.STATE_TABLE_ONLY # no change
S
Steven Li 已提交
658 659 660
        # else: # did not drop table, did not insert data, did not read successfully, that is impossible
        #     raise RuntimeError("Unexpected no-success scenarios")
        # TODO: need to revamp!!
661 662 663 664 665 666 667 668 669 670 671

class StateHasData(AnyState):
    def getInfo(self):
        return [
            self.STATE_HAS_DATA,
            False, True,
            False, True,
            True, True,
        ]

    def verifyTasksToState(self, tasks, newState):
S
Steven Li 已提交
672 673 674
        if ( newState.equals(AnyState.STATE_EMPTY) ):
            self.hasSuccess(tasks, DropDbTask)
            self.assertAtMostOneSuccess(tasks, DropDbTask) # TODO: dicy
S
Steven Li 已提交
675 676 677
        elif ( newState.equals(AnyState.STATE_DB_ONLY) ): # in DB only
            if ( not self.hasTask(tasks, CreateDbTask)): # without a create_db task
                self.assertNoTask(tasks, DropDbTask) # we must have drop_db task
678
            self.hasSuccess(tasks, DropFixedSuperTableTask)
679
            # self.assertAtMostOneSuccess(tasks, DropFixedSuperTableTask) # TODO: dicy
S
Steven Li 已提交
680 681
        elif ( newState.equals(AnyState.STATE_TABLE_ONLY) ): # data deleted
            self.assertNoTask(tasks, DropDbTask)
682
            self.assertNoTask(tasks, DropFixedSuperTableTask)
S
Steven Li 已提交
683 684 685 686
            self.assertNoTask(tasks, AddFixedDataTask)
            # self.hasSuccess(tasks, DeleteDataTasks)
        else:
            self.assertNoTask(tasks, DropDbTask)
687
            self.assertNoTask(tasks, DropFixedSuperTableTask)
S
Steven Li 已提交
688 689
            self.assertIfExistThenSuccess(tasks, ReadFixedDataTask)

690

691 692 693
# State of the database as we believe it to be
class DbState():
    
694
    def __init__(self, resetDb = True):
S
Steven Li 已提交
695
        self.tableNumQueue = LinearQueue()
696
        self._lastTick = self.setupLastTick() # datetime.datetime(2019, 1, 1) # initial date time tick
697 698
        self._lastInt  = 0 # next one is initial integer 
        self._lock = threading.RLock()
699

700
        self._state = StateInvalid() # starting state
701
        self._stateWeights = [1,3,5,10] # indexed with value of STATE_EMPTY, STATE_DB_ONLY, etc.
702
        
703 704
        # self.openDbServerConnection()
        self._dbConn = DbConn()
705 706 707 708 709 710 711 712 713 714
        try:
            self._dbConn.open() # may throw taos.error.ProgrammingError: disconnected
        except taos.error.ProgrammingError as err:
            # print("Error type: {}, msg: {}, value: {}".format(type(err), err.msg, err))
            if ( err.msg == 'disconnected' ): # cannot open DB connection
                print("Cannot establish DB connection, please re-run script without parameter, and follow the instructions.")
                sys.exit()
            else:
                raise            
        except:
S
Steven Li 已提交
715
            print("[=] Unexpected exception")
716
            raise        
717 718 719 720

        if resetDb :
            self._dbConn.resetDb() # drop and recreate DB            
        self._state = self._findCurrentState()
721

722 723 724
    def getDbConn(self):
        return self._dbConn

725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742
    def getState(self):
        return self._state

    # We aim to create a starting time tick, such that, whenever we run our test here once
    # We should be able to safely create 100,000 records, which will not have any repeated time stamp
    # when we re-run the test in 3 minutes (180 seconds), basically we should expand time duration
    # by a factor of 500.
    # TODO: what if it goes beyond 10 years into the future
    def setupLastTick(self):
        t1 = datetime.datetime(2020, 5, 30)
        t2 = datetime.datetime.now()
        elSec = t2.timestamp() - t1.timestamp()
        # print("elSec = {}".format(elSec))
        t3 = datetime.datetime(2012, 1, 1) # default "keep" is 10 years
        t4 = datetime.datetime.fromtimestamp( t3.timestamp() + elSec * 500) # see explanation above
        logger.info("Setting up TICKS to start from: {}".format(t4))
        return t4

S
Steven Li 已提交
743 744 745
    def pickAndAllocateTable(self): # pick any table, and "use" it
        return self.tableNumQueue.pickAndAllocate()

746 747 748 749 750
    def addTable(self):
        with self._lock:
            tIndex = self.tableNumQueue.push()
        return tIndex

751 752
    def getFixedSuperTableName(self):
        return "fs_table"
753

S
Steven Li 已提交
754 755 756
    def releaseTable(self, i): # return the table back, so others can use it
        self.tableNumQueue.release(i)

757
    def getNextTick(self):
758 759 760
        with self._lock: # prevent duplicate tick
            self._lastTick += datetime.timedelta(0, 1) # add one second to it
            return self._lastTick
761 762

    def getNextInt(self):
763 764 765
        with self._lock:
            self._lastInt += 1
            return self._lastInt
766 767

    def getNextBinary(self):
768
        return "Beijing_Shanghai_Los_Angeles_New_York_San_Francisco_Chicago_Beijing_Shanghai_Los_Angeles_New_York_San_Francisco_Chicago_{}".format(self.getNextInt())
769 770 771

    def getNextFloat(self):
        return 0.9 + self.getNextInt()
772
    
S
Steven Li 已提交
773
    def getTableNameToDelete(self):
774
        tblNum = self.tableNumQueue.pop() # TODO: race condition!
775 776 777
        if ( not tblNum ): # maybe false
            return False
        
S
Steven Li 已提交
778 779
        return "table_{}".format(tblNum)

780 781 782
    def cleanUp(self):
        self._dbConn.close()      

S
Steven Li 已提交
783 784
    # May be slow, use cautionsly...
    def getTaskTypesAtState(self):        
785
        allTaskClasses = StateTransitionTask.__subclasses__() # all state transition tasks
S
Steven Li 已提交
786
        firstTaskTypes = []
787
        for tc in allTaskClasses:
S
Steven Li 已提交
788
            # t = tc(self) # create task object            
789
            if tc.canBeginFrom(self._state):
S
Steven Li 已提交
790 791 792 793 794 795 796 797
                firstTaskTypes.append(tc)
        # now we have all the tasks that can begin directly from the current state, let's figure out the INDIRECT ones
        taskTypes = firstTaskTypes.copy() # have to have these
        for task1 in firstTaskTypes: # each task type gathered so far
            endState = task1.getEndState() # figure the end state
            if endState == None:
                continue
            for tc in allTaskClasses: # what task can further begin from there?
798
                if tc.canBeginFrom(endState) and (tc not in firstTaskTypes):
S
Steven Li 已提交
799 800
                    taskTypes.append(tc) # gather it

801
        if len(taskTypes) <= 0:
802 803
            raise RuntimeError("No suitable task types found for state: {}".format(self._state))   
        logger.debug("[OPS] Tasks found for state {}: {}".format(self._state, taskTypes))     
804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829
        return taskTypes

        # tasks.append(ReadFixedDataTask(self)) # always for everybody
        # if ( self._state == self.STATE_EMPTY ):
        #     tasks.append(CreateDbTask(self))
        #     tasks.append(CreateFixedTableTask(self))
        # elif ( self._state == self.STATE_DB_ONLY ):
        #     tasks.append(DropDbTask(self))
        #     tasks.append(CreateFixedTableTask(self))
        #     tasks.append(AddFixedDataTask(self))
        # elif ( self._state == self.STATE_TABLE_ONLY ):
        #     tasks.append(DropFixedTableTask(self))
        #     tasks.append(AddFixedDataTask(self))
        # elif ( self._state == self.STATE_HAS_DATA ) : # same as above. TODO: adjust
        #     tasks.append(DropFixedTableTask(self))
        #     tasks.append(AddFixedDataTask(self))
        # else:
        #     raise RuntimeError("Unexpected DbState state: {}".format(self._state))
        # return tasks

    def pickTaskType(self):
        taskTypes = self.getTaskTypesAtState() # all the task types we can choose from at curent state
        weights = []
        for tt in taskTypes:
            endState = tt.getEndState()
            if endState != None :
S
Steven Li 已提交
830
                weights.append(self._stateWeights[endState.getValIndex()]) # TODO: change to a method
831 832 833
            else:
                weights.append(10) # read data task, default to 10: TODO: change to a constant
        i = self._weighted_choice_sub(weights)
834
        # logger.debug(" (weighted random:{}/{}) ".format(i, len(taskTypes)))        
835 836 837 838 839 840 841 842 843
        return taskTypes[i]

    def _weighted_choice_sub(self, weights): # ref: https://eli.thegreenplace.net/2010/01/22/weighted-random-generation-in-python/
        rnd = random.random() * sum(weights) # TODO: use our dice to ensure it being determinstic?
        for i, w in enumerate(weights):
            rnd -= w
            if rnd < 0:
                return i

844 845
    def _findCurrentState(self):
        dbc = self._dbConn
S
Steven Li 已提交
846
        ts = time.time()
847
        if dbc.query("show databases") == 0 : # no database?!
S
Steven Li 已提交
848
            # logger.debug("Found EMPTY state")
S
Steven Li 已提交
849
            logger.debug("[STT] empty database found, between {} and {}".format(ts, time.time()))
850 851 852
            return StateEmpty()
        dbc.execute("use db") # did not do this when openning connection
        if dbc.query("show tables") == 0 : # no tables
S
Steven Li 已提交
853
            # logger.debug("Found DB ONLY state")
S
Steven Li 已提交
854
            logger.debug("[STT] DB_ONLY found, between {} and {}".format(ts, time.time()))
855
            return StateDbOnly()
856
        if dbc.query("SELECT * FROM db.{}".format(self.getFixedSuperTableName()) ) == 0 : # no data
S
Steven Li 已提交
857
            # logger.debug("Found TABLE_ONLY state")
858 859
            logger.debug("[STT] SUPER_TABLE_ONLY found, between {} and {}".format(ts, time.time()))
            return StateSuperTableOnly()
860
        else:
S
Steven Li 已提交
861
            # logger.debug("Found HAS_DATA state")
S
Steven Li 已提交
862
            logger.debug("[STT] HAS_DATA found, between {} and {}".format(ts, time.time()))
863
            return StateHasData()
864
    
865 866 867
    def transition(self, tasks):
        if ( len(tasks) == 0 ): # before 1st step, or otherwise empty
            return # do nothing
868

869
        self._dbConn.execute("show dnodes") # this should show up in the server log, separating steps
870

871 872 873 874
        # Generic Checks, first based on the start state
        if self._state.canCreateDb():
            self._state.assertIfExistThenSuccess(tasks, CreateDbTask)
            # self.assertAtMostOneSuccess(tasks, CreateDbTask) # not really, in case of multiple creation and drops
875

876 877 878
        if self._state.canDropDb():
            self._state.assertIfExistThenSuccess(tasks, DropDbTask)
            # self.assertAtMostOneSuccess(tasks, DropDbTask) # not really in case of drop-create-drop
879

880 881 882
        # if self._state.canCreateFixedTable():
            # self.assertIfExistThenSuccess(tasks, CreateFixedTableTask) # Not true, DB may be dropped
            # self.assertAtMostOneSuccess(tasks, CreateFixedTableTask) # not really, in case of create-drop-create
883

884 885 886
        # if self._state.canDropFixedTable():
            # self.assertIfExistThenSuccess(tasks, DropFixedTableTask) # Not True, the whole DB may be dropped
            # self.assertAtMostOneSuccess(tasks, DropFixedTableTask) # not really in case of drop-create-drop
887

888 889
        # if self._state.canAddData():
        #     self.assertIfExistThenSuccess(tasks, AddFixedDataTask)  # not true actually
890

891 892
        # if self._state.canReadData():
            # Nothing for sure
893

894
        newState = self._findCurrentState()
S
Steven Li 已提交
895
        logger.debug("[STT] New DB state determined: {}".format(newState))
896 897
        self._state.verifyTasksToState(tasks, newState) # can old state move to new state through the tasks?
        self._state = newState
898

899 900 901 902
class TaskExecutor():
    def __init__(self, curStep):
        self._curStep = curStep

903 904 905
    def getCurStep(self):
        return self._curStep

906 907
    def execute(self, task: Task, wt: WorkerThread): # execute a task on a thread
        task.execute(wt)
908

909 910
    # def logInfo(self, msg):
    #     logger.info("    T[{}.x]: ".format(self._curStep) + msg)
911

912 913
    # def logDebug(self, msg):
    #     logger.debug("    T[{}.x]: ".format(self._curStep) + msg)
914

S
Steven Li 已提交
915
class Task():
916 917 918 919
    taskSn = 100

    @classmethod
    def allocTaskNum(cls):
S
Steven Li 已提交
920 921 922
        Task.taskSn += 1 # IMPORTANT: cannot use cls.taskSn, since each sub class will have a copy
        # logger.debug("Allocating taskSN: {}".format(Task.taskSn))
        return Task.taskSn
923

S
Steven Li 已提交
924
    def __init__(self, dbState: DbState, execStats: ExecutionStats):        
925
        self._dbState = dbState
926
        self._workerThread = None 
927
        self._err = None
928
        self._curStep = None
929
        self._numRows = None # Number of rows affected
930 931 932

        # Assign an incremental task serial number        
        self._taskNum = self.allocTaskNum()
S
Steven Li 已提交
933
        # logger.debug("Creating new task {}...".format(self._taskNum))
934

935 936
        self._execStats = execStats

937 938
    def isSuccess(self):
        return self._err == None
939

940 941
    def clone(self): # TODO: why do we need this again?
        newTask = self.__class__(self._dbState, self._execStats)
942 943 944
        return newTask

    def logDebug(self, msg):
S
Steven Li 已提交
945
        self._workerThread.logDebug("Step[{}.{}] {}".format(self._curStep, self._taskNum, msg))
946 947

    def logInfo(self, msg):
S
Steven Li 已提交
948
        self._workerThread.logInfo("Step[{}.{}] {}".format(self._curStep, self._taskNum, msg))
949

950
    def _executeInternal(self, te: TaskExecutor, wt: WorkerThread):
951
        raise RuntimeError("To be implemeted by child classes, class name: {}".format(self.__class__.__name__))
952

953 954
    def execute(self, wt: WorkerThread):
        wt.verifyThreadSelf()
955
        self._workerThread = wt # type: ignore
956 957

        te = wt.getTaskExecutor()
958 959
        self._curStep = te.getCurStep()
        self.logDebug("[-] executing task {}...".format(self.__class__.__name__))
960 961

        self._err = None
962
        self._execStats.beginTaskType(self.__class__.__name__) # mark beginning
963 964 965
        try:
            self._executeInternal(te, wt) # TODO: no return value?
        except taos.error.ProgrammingError as err:
S
Steven Li 已提交
966
            self.logDebug("[=] Taos library exception: errno={}, msg: {}".format(err.errno, err))
967 968
            self._err = err
        except:
S
Steven Li 已提交
969
            self.logDebug("[=] Unexpected exception")
970
            raise
971
        self._execStats.endTaskType(self.__class__.__name__, self.isSuccess())
972
        
973 974
        self.logDebug("[X] task execution completed, {}, status: {}".format(self.__class__.__name__, "Success" if self.isSuccess() else "Failure"))        
        self._execStats.incExecCount(self.__class__.__name__, self.isSuccess()) # TODO: merge with above.
S
Steven Li 已提交
975

976
    def execSql(self, sql):
977
        return self._dbState.execute(sql)
978

979
                  
980
class ExecutionStats:
981 982 983 984 985
    def __init__(self):
        self._execTimes: Dict[str, [int, int]] = {} # total/success times for a task
        self._tasksInProgress = 0
        self._lock = threading.Lock()
        self._firstTaskStartTime = None
986 987
        self._execStartTime = None
        self._elapsedTime = 0.0 # total elapsed time
988 989
        self._accRunTime = 0.0 # accumulated run time

990 991 992 993 994 995 996 997 998
        self._failed = False
        self._failureReason = None

    def startExec(self):
        self._execStartTime = time.time()

    def endExec(self):
        self._elapsedTime = time.time() - self._execStartTime

999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019
    def incExecCount(self, klassName, isSuccess): # TODO: add a lock here
        if klassName not in self._execTimes:
            self._execTimes[klassName] = [0, 0]
        t = self._execTimes[klassName] # tuple for the data
        t[0] += 1 # index 0 has the "total" execution times
        if isSuccess:
            t[1] += 1 # index 1 has the "success" execution times

    def beginTaskType(self, klassName):
        with self._lock:
            if self._tasksInProgress == 0 : # starting a new round
                self._firstTaskStartTime = time.time() # I am now the first task
            self._tasksInProgress += 1

    def endTaskType(self, klassName, isSuccess):
        with self._lock:
            self._tasksInProgress -= 1
            if self._tasksInProgress == 0 : # all tasks have stopped
                self._accRunTime += (time.time() - self._firstTaskStartTime)
                self._firstTaskStartTime = None

1020 1021 1022 1023
    def registerFailure(self, reason):
        self._failed = True
        self._failureReason = reason

1024
    def logStats(self):
1025 1026 1027 1028
        logger.info("----------------------------------------------------------------------")
        logger.info("| Crash_Gen test {}, with the following stats:".
            format("FAILED (reason: {})".format(self._failureReason) if self._failed else "SUCCEEDED"))
        logger.info("| Task Execution Times (success/total):")
1029 1030
        execTimesAny = 0
        for k, n in self._execTimes.items():            
1031
            execTimesAny += n[0]
1032
            logger.info("|    {0:<24}: {1}/{2}".format(k,n[1],n[0]))
1033
                
1034 1035 1036 1037 1038 1039 1040
        logger.info("| Total Tasks Executed (success or not): {} ".format(execTimesAny))
        logger.info("| Total Tasks In Progress at End: {}".format(self._tasksInProgress))
        logger.info("| Total Task Busy Time (elapsed time when any task is in progress): {:.3f} seconds".format(self._accRunTime))
        logger.info("| Average Per-Task Execution Time: {:.3f} seconds".format(self._accRunTime/execTimesAny))
        logger.info("| Total Elapsed Time (from wall clock): {:.3f} seconds".format(self._elapsedTime))
        logger.info("----------------------------------------------------------------------")
        
1041 1042 1043 1044 1045 1046 1047 1048 1049 1050


class StateTransitionTask(Task):
    # @classmethod
    # def getAllTaskClasses(cls): # static
    #     return cls.__subclasses__()
    @classmethod
    def getInfo(cls): # each sub class should supply their own information
        raise RuntimeError("Overriding method expected")

1051 1052 1053
    # @classmethod
    # def getBeginStates(cls):
    #     return cls.getInfo()[0]
1054 1055

    @classmethod
S
Steven Li 已提交
1056
    def getEndState(cls): # returning the class name
1057
        return cls.getInfo()[0]
1058 1059

    @classmethod
1060 1061 1062
    def canBeginFrom(cls, state: AnyState):
        # return state.getValue() in cls.getBeginStates()
        raise RuntimeError("must be overriden")
1063 1064 1065 1066 1067 1068 1069 1070 1071 1072

    def execute(self, wt: WorkerThread):
        super().execute(wt)
        


class CreateDbTask(StateTransitionTask):
    @classmethod
    def getInfo(cls):
        return [
1073
            # [AnyState.STATE_EMPTY], # can begin from
S
Steven Li 已提交
1074
            StateDbOnly() # end state
1075 1076
        ]

1077 1078 1079 1080
    @classmethod
    def canBeginFrom(cls, state: AnyState):
        return state.canCreateDb()

1081
    def _executeInternal(self, te: TaskExecutor, wt: WorkerThread):
1082
        wt.execSql("create database db")       
1083

1084 1085 1086 1087
class DropDbTask(StateTransitionTask):
    @classmethod
    def getInfo(cls):
        return [
1088
            # [AnyState.STATE_DB_ONLY, AnyState.STATE_TABLE_ONLY, AnyState.STATE_HAS_DATA],
S
Steven Li 已提交
1089
            StateEmpty()
1090 1091
        ]

1092 1093 1094 1095
    @classmethod
    def canBeginFrom(cls, state: AnyState):
        return state.canDropDb()

1096 1097
    def _executeInternal(self, te: TaskExecutor, wt: WorkerThread):
        wt.execSql("drop database db")
S
Steven Li 已提交
1098
        logger.debug("[OPS] database dropped at {}".format(time.time()))
1099

1100
class CreateFixedSuperTableTask(StateTransitionTask):
1101 1102 1103
    @classmethod
    def getInfo(cls):
        return [
1104
            # [AnyState.STATE_DB_ONLY],
1105
            StateSuperTableOnly()
1106
        ]
1107

1108 1109
    @classmethod
    def canBeginFrom(cls, state: AnyState):
1110
        return state.canCreateFixedSuperTable()
1111

1112
    def _executeInternal(self, te: TaskExecutor, wt: WorkerThread):
1113
        tblName = self._dbState.getFixedSuperTableName()        
1114
        wt.execSql("create table db.{} (ts timestamp, speed int) tags (b binary(200), f float) ".format(tblName))
1115 1116
        # No need to create the regular tables, INSERT will do that automatically

S
Steven Li 已提交
1117

1118 1119 1120 1121
class ReadFixedDataTask(StateTransitionTask):
    @classmethod
    def getInfo(cls):
        return [
1122
            # [AnyState.STATE_TABLE_ONLY, AnyState.STATE_HAS_DATA],
1123 1124 1125
            None # meaning doesn't affect state
        ]

1126 1127 1128 1129
    @classmethod
    def canBeginFrom(cls, state: AnyState):
        return state.canReadData()

1130
    def _executeInternal(self, te: TaskExecutor, wt: WorkerThread):
1131 1132 1133 1134 1135 1136 1137 1138
        sTbName = self._dbState.getFixedSuperTableName()        
        dbc = wt.getDbConn()
        dbc.query("select TBNAME from db.{}".format(sTbName)) # TODO: analyze result set later
        rTables = dbc.getQueryResult()
        # print("rTables[0] = {}, type = {}".format(rTables[0], type(rTables[0])))
        for rTbName in rTables : # regular tables
            dbc.query("select * from db.{}".format(rTbName[0])) # TODO: check success failure

1139 1140
        # tdSql.query(" cars where tbname in ('carzero', 'carone')")

1141
class DropFixedSuperTableTask(StateTransitionTask):
1142 1143 1144
    @classmethod
    def getInfo(cls):
        return [
1145
            # [AnyState.STATE_TABLE_ONLY, AnyState.STATE_HAS_DATA],
S
Steven Li 已提交
1146
            StateDbOnly() # meaning doesn't affect state
1147 1148
        ]

1149 1150
    @classmethod
    def canBeginFrom(cls, state: AnyState):
1151
        return state.canDropFixedSuperTable()
1152

1153
    def _executeInternal(self, te: TaskExecutor, wt: WorkerThread):
1154
        tblName = self._dbState.getFixedSuperTableName()        
1155 1156 1157 1158 1159 1160
        wt.execSql("drop table db.{}".format(tblName))

class AddFixedDataTask(StateTransitionTask):
    @classmethod
    def getInfo(cls):
        return [
1161
            # [AnyState.STATE_TABLE_ONLY, AnyState.STATE_HAS_DATA],
S
Steven Li 已提交
1162
            StateHasData()
1163
        ]
1164 1165 1166 1167

    @classmethod
    def canBeginFrom(cls, state: AnyState):
        return state.canAddData()
1168 1169 1170
        
    def _executeInternal(self, te: TaskExecutor, wt: WorkerThread):
        ds = self._dbState
1171 1172
        wt.execSql("use db") # TODO: seems to be an INSERT bug to require this
        for i in range(10): # 0 to 9
1173
            for j in range(10) :
1174 1175 1176 1177 1178 1179
                sql = "insert into db.reg_table_{} using {} tags ('{}', {}) values ('{}', {});".format(
                    i, 
                    ds.getFixedSuperTableName(), 
                    ds.getNextBinary(), ds.getNextFloat(),
                    ds.getNextTick(), ds.getNextInt())
                wt.execSql(sql) 
1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191


#---------- Non State-Transition Related Tasks ----------#

class CreateTableTask(Task):    
    def _executeInternal(self, te: TaskExecutor, wt: WorkerThread):
        tIndex = self._dbState.addTable()
        self.logDebug("Creating a table {} ...".format(tIndex))
        wt.execSql("create table db.table_{} (ts timestamp, speed int)".format(tIndex))
        self.logDebug("Table {} created.".format(tIndex))
        self._dbState.releaseTable(tIndex)

S
Steven Li 已提交
1192
class DropTableTask(Task):
1193
    def _executeInternal(self, te: TaskExecutor, wt: WorkerThread):
1194
        tableName = self._dbState.getTableNameToDelete()
S
Steven Li 已提交
1195
        if ( not tableName ): # May be "False"
1196
            self.logInfo("Cannot generate a table to delete, skipping...")
S
Steven Li 已提交
1197
            return
1198
        self.logInfo("Dropping a table db.{} ...".format(tableName))
1199
        wt.execSql("drop table db.{}".format(tableName))
1200
        
1201

S
Steven Li 已提交
1202 1203

class AddDataTask(Task):
1204
    def _executeInternal(self, te: TaskExecutor, wt: WorkerThread):
1205
        ds = self._dbState
1206
        self.logInfo("Adding some data... numQueue={}".format(ds.tableNumQueue.toText()))
1207 1208
        tIndex = ds.pickAndAllocateTable()
        if ( tIndex == None ):
1209
            self.logInfo("No table found to add data, skipping...")
1210
            return
1211
        sql = "insert into db.table_{} values ('{}', {});".format(tIndex, ds.getNextTick(), ds.getNextInt())
1212
        self.logDebug("[SQL] Executing SQL: {}".format(sql))
1213 1214
        wt.execSql(sql) 
        ds.releaseTable(tIndex)
1215
        self.logDebug("[OPS] Finished adding data")
S
Steven Li 已提交
1216

1217

S
Steven Li 已提交
1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239
# Deterministic random number generator
class Dice():
    seeded = False # static, uninitialized

    @classmethod
    def seed(cls, s): # static
        if (cls.seeded):
            raise RuntimeError("Cannot seed the random generator more than once")
        cls.verifyRNG()
        random.seed(s)
        cls.seeded = True  # TODO: protect against multi-threading

    @classmethod
    def verifyRNG(cls): # Verify that the RNG is determinstic
        random.seed(0)
        x1 = random.randrange(0, 1000)
        x2 = random.randrange(0, 1000)
        x3 = random.randrange(0, 1000)
        if ( x1 != 864 or x2!=394 or x3!=776 ):
            raise RuntimeError("System RNG is not deterministic")

    @classmethod
1240 1241
    def throw(cls, stop): # get 0 to stop-1
        return cls.throwRange(0, stop)
S
Steven Li 已提交
1242 1243

    @classmethod
1244
    def throwRange(cls, start, stop): # up to stop-1
S
Steven Li 已提交
1245 1246
        if ( not cls.seeded ):
            raise RuntimeError("Cannot throw dice before seeding it")
1247
        return random.randrange(start, stop)
S
Steven Li 已提交
1248 1249 1250


# Anyone needing to carry out work should simply come here
1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272
# class WorkDispatcher():
#     def __init__(self, dbState):
#         # self.totalNumMethods = 2
#         self.tasks = [
#             # CreateTableTask(dbState), # Obsolete
#             # DropTableTask(dbState),
#             # AddDataTask(dbState),
#         ]

#     def throwDice(self):
#         max = len(self.tasks) - 1 
#         dRes = random.randint(0, max)
#         # logger.debug("Threw the dice in range [{},{}], and got: {}".format(0,max,dRes))
#         return dRes

#     def pickTask(self):
#         dice = self.throwDice()
#         return self.tasks[dice]

#     def doWork(self, workerThread):
#         task = self.pickTask()
#         task.execute(workerThread)
S
Steven Li 已提交
1273

S
Steven Li 已提交
1274 1275
class LoggingFilter(logging.Filter):
    def filter(self, record: logging.LogRecord):
S
Steven Li 已提交
1276 1277 1278
        if ( record.levelno >= logging.INFO ) :
            return True # info or above always log

S
Steven Li 已提交
1279 1280 1281
        msg = record.msg
        # print("type = {}, value={}".format(type(msg), msg))
        # sys.exit()
S
Steven Li 已提交
1282 1283 1284 1285 1286

        # Commenting out below to adjust...

        # if msg.startswith("[TRD]"):
        #     return False
S
Steven Li 已提交
1287 1288 1289 1290
        return True

        

1291
def main():
1292
    # Super cool Python argument library: https://docs.python.org/3/library/argparse.html
1293 1294 1295 1296 1297 1298 1299 1300 1301
    parser = argparse.ArgumentParser(
        formatter_class=argparse.RawDescriptionHelpFormatter,
        description=textwrap.dedent('''\
            TDengine Auto Crash Generator (PLEASE NOTICE the Prerequisites Below)
            ---------------------------------------------------------------------
            1. You build TDengine in the top level ./build directory, as described in offical docs
            2. You run the server there before this script: ./build/bin/taosd -c test/cfg

            '''))
1302 1303 1304 1305
    parser.add_argument('-p', '--per-thread-db-connection', action='store_true',                        
                        help='Use a single shared db connection (default: false)')
    parser.add_argument('-d', '--debug', action='store_true',                        
                        help='Turn on DEBUG mode for more logging (default: false)')
1306 1307 1308 1309
    parser.add_argument('-s', '--max-steps', action='store', default=100, type=int,
                        help='Maximum number of steps to run (default: 100)')
    parser.add_argument('-t', '--num-threads', action='store', default=10, type=int,
                        help='Number of threads to run (default: 10)')
1310

1311
    global gConfig
1312
    gConfig = parser.parse_args()
1313 1314 1315
    if len(sys.argv) == 1:
        parser.print_help()
        sys.exit()
1316

1317
    global logger
1318
    logger = logging.getLogger('CrashGen')
S
Steven Li 已提交
1319
    logger.addFilter(LoggingFilter())
1320 1321
    if ( gConfig.debug ):
        logger.setLevel(logging.DEBUG) # default seems to be INFO        
S
Steven Li 已提交
1322 1323
    else:
        logger.setLevel(logging.INFO)
S
Steven Li 已提交
1324 1325 1326
    ch = logging.StreamHandler()
    logger.addHandler(ch)

1327 1328 1329
    # resetDb = False # DEBUG only
    # dbState = DbState(resetDb)  # DBEUG only!
    dbState = DbState() # Regular function
1330 1331
    Dice.seed(0) # initial seeding of dice
    tc = ThreadCoordinator(
1332
        ThreadPool(dbState, gConfig.num_threads, gConfig.max_steps, 0), 
1333
        # WorkDispatcher(dbState), # Obsolete?
1334
        dbState
1335
        )
S
Steven Li 已提交
1336

1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371
    # # Hack to exercise reading from disk, imcreasing coverage. TODO: fix
    # dbc = dbState.getDbConn()
    # sTbName = dbState.getFixedSuperTableName()   
    # dbc.execute("create database if not exists db")
    # if not dbState.getState().equals(StateEmpty()):
    #     dbc.execute("use db")     

    # rTables = None
    # try: # the super table may not exist
    #     sql = "select TBNAME from db.{}".format(sTbName)
    #     logger.info("Finding out tables in super table: {}".format(sql))
    #     dbc.query(sql) # TODO: analyze result set later
    #     logger.info("Fetching result")
    #     rTables = dbc.getQueryResult()
    #     logger.info("Result: {}".format(rTables))
    # except taos.error.ProgrammingError as err:
    #     logger.info("Initial Super table OPS error: {}".format(err))
    
    # # sys.exit()
    # if ( not rTables == None):
    #     # print("rTables[0] = {}, type = {}".format(rTables[0], type(rTables[0])))
    #     try:
    #         for rTbName in rTables : # regular tables
    #             ds = dbState
    #             logger.info("Inserting into table: {}".format(rTbName[0]))
    #             sql = "insert into db.{} values ('{}', {});".format(
    #                 rTbName[0],                    
    #                 ds.getNextTick(), ds.getNextInt())
    #             dbc.execute(sql)
    #         for rTbName in rTables : # regular tables        
    #             dbc.query("select * from db.{}".format(rTbName[0])) # TODO: check success failure
    #         logger.info("Initial READING operation is successful")       
    #     except taos.error.ProgrammingError as err:
    #         logger.info("Initial WRITE/READ error: {}".format(err))   
    
1372 1373
    

S
Steven Li 已提交
1374 1375 1376 1377 1378
    # Sandbox testing code
    # dbc = dbState.getDbConn()
    # while True:
    #     rows = dbc.query("show databases") 
    #     print("Rows: {}, time={}".format(rows, time.time()))
1379
    
1380
    tc.run()
1381 1382 1383
    tc.logStats()
    dbState.cleanUp()    
    
S
Steven Li 已提交
1384
    # logger.info("Crash_Gen execution finished")
1385 1386 1387

if __name__ == "__main__":
    main()