1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708
|
import datetime
import logging
import multiprocessing.queues as mpq
import os
import queue
import threading
import time
from typing import Any, Dict, List, Optional, Set, Tuple, TypeVar, cast
import typeguard
from parsl.dataflow.states import States
from parsl.errors import OptionalModuleMissing
from parsl.log_utils import set_file_logger
from parsl.monitoring.message_type import MessageType
from parsl.monitoring.types import MonitoringMessage, TaggedMonitoringMessage
from parsl.process_loggers import wrap_with_logs
from parsl.utils import setproctitle
logger = logging.getLogger("database_manager")
X = TypeVar('X')
try:
import sqlalchemy as sa
from sqlalchemy import (
BigInteger,
Boolean,
Column,
DateTime,
Float,
Integer,
PrimaryKeyConstraint,
Table,
Text,
)
from sqlalchemy.orm import Mapper, declarative_base, mapperlib, sessionmaker
except ImportError:
_sqlalchemy_enabled = False
else:
_sqlalchemy_enabled = True
WORKFLOW = 'workflow' # Workflow table includes workflow metadata
TASK = 'task' # Task table includes task metadata
TRY = 'try' # Try table includes information about each attempt to run a task
STATUS = 'status' # Status table includes task status
RESOURCE = 'resource' # Resource table includes task resource utilization
NODE = 'node' # Node table include node info
BLOCK = 'block' # Block table include the status for block polling
class Database:
if not _sqlalchemy_enabled:
raise OptionalModuleMissing(['sqlalchemy'],
("Monitoring requires the sqlalchemy library."
" Install monitoring dependencies with: pip install 'parsl[monitoring]'"))
Base = declarative_base()
def __init__(self,
url: str = 'sqlite:///runinfomonitoring.db',
):
self.eng = sa.create_engine(url)
self.meta = self.Base.metadata
# TODO: this code wants a read lock on the sqlite3 database, and fails if it cannot
# - for example, if someone else is querying the database at the point that the
# monitoring system is initialized. See PR #1917 for related locked-for-read fixes
# elsewhere in this file.
self.meta.create_all(self.eng)
self.meta.reflect(bind=self.eng)
Session = sessionmaker(bind=self.eng)
self.session = Session()
def _get_mapper(self, table_obj: Table) -> Mapper:
all_mappers: Set[Mapper] = set()
for mapper_registry in mapperlib._all_registries():
all_mappers.update(mapper_registry.mappers)
mapper_gen = (
mapper for mapper in all_mappers
if table_obj in mapper.tables
)
try:
mapper = next(mapper_gen)
second_mapper = next(mapper_gen, False)
except StopIteration:
raise ValueError(f"Could not get mapper for table {table_obj}")
if second_mapper:
raise ValueError(f"Multiple mappers for table {table_obj}")
return mapper
def update(self, *, table: str, columns: List[str], messages: List[MonitoringMessage]) -> None:
table_obj = self.meta.tables[table]
mappings = self._generate_mappings(table_obj, columns=columns,
messages=messages)
mapper = self._get_mapper(table_obj)
self.session.bulk_update_mappings(mapper, mappings)
self.session.commit()
def insert(self, *, table: str, messages: List[MonitoringMessage]) -> None:
table_obj = self.meta.tables[table]
mappings = self._generate_mappings(table_obj, messages=messages)
mapper = self._get_mapper(table_obj)
self.session.bulk_insert_mappings(mapper, mappings)
self.session.commit()
def rollback(self) -> None:
self.session.rollback()
def _generate_mappings(
self,
table: Table,
columns: Optional[List[str]] = None,
messages: List[MonitoringMessage] = [],
) -> List[Dict[str, Any]]:
mappings = []
for msg in messages:
m = {}
if columns is None:
columns = table.c.keys()
for column in columns:
m[column] = msg.get(column, None)
mappings.append(m)
return mappings
class Workflow(Base):
__tablename__ = WORKFLOW
run_id = Column(Text, nullable=False, primary_key=True)
workflow_name = Column(Text, nullable=True)
workflow_version = Column(Text, nullable=True)
time_began = Column(DateTime, nullable=False)
time_completed = Column(DateTime, nullable=True)
host = Column(Text, nullable=False)
user = Column(Text, nullable=False)
rundir = Column(Text, nullable=False)
tasks_failed_count = Column(Integer, nullable=False)
tasks_completed_count = Column(Integer, nullable=False)
class Status(Base):
__tablename__ = STATUS
task_id = Column(Integer, nullable=False)
task_status_name = Column(Text, nullable=False)
timestamp = Column(DateTime, nullable=False)
run_id = Column(Text, sa.ForeignKey('workflow.run_id'), nullable=False)
try_id = Column('try_id', Integer, nullable=False)
__table_args__ = (
PrimaryKeyConstraint('task_id', 'run_id',
'task_status_name', 'timestamp'),
)
class Task(Base):
__tablename__ = TASK
task_id = Column('task_id', Integer, nullable=False)
run_id = Column('run_id', Text, nullable=False)
task_depends = Column('task_depends', Text, nullable=True)
task_func_name = Column('task_func_name', Text, nullable=False)
task_memoize = Column('task_memoize', Text, nullable=False)
task_hashsum = Column('task_hashsum', Text, nullable=True, index=True)
task_inputs = Column('task_inputs', Text, nullable=True)
task_outputs = Column('task_outputs', Text, nullable=True)
task_stdin = Column('task_stdin', Text, nullable=True)
task_stdout = Column('task_stdout', Text, nullable=True)
task_stderr = Column('task_stderr', Text, nullable=True)
task_time_invoked = Column(
'task_time_invoked', DateTime, nullable=True)
task_time_returned = Column(
'task_time_returned', DateTime, nullable=True)
task_fail_count = Column('task_fail_count', Integer, nullable=False)
task_fail_cost = Column('task_fail_cost', Float, nullable=False)
__table_args__ = (
PrimaryKeyConstraint('task_id', 'run_id'),
)
class Try(Base):
__tablename__ = TRY
try_id = Column('try_id', Integer, nullable=False)
task_id = Column('task_id', Integer, nullable=False)
run_id = Column('run_id', Text, nullable=False)
block_id = Column('block_id', Text, nullable=True)
hostname = Column('hostname', Text, nullable=True)
task_executor = Column('task_executor', Text, nullable=False)
task_try_time_launched = Column(
'task_try_time_launched', DateTime, nullable=True)
task_try_time_running = Column(
'task_try_time_running', DateTime, nullable=True)
task_try_time_returned = Column(
'task_try_time_returned', DateTime, nullable=True)
task_fail_history = Column('task_fail_history', Text, nullable=True)
task_joins = Column('task_joins', Text, nullable=True)
__table_args__ = (
PrimaryKeyConstraint('try_id', 'task_id', 'run_id'),
)
class Node(Base):
__tablename__ = NODE
id = Column('id', Integer, nullable=False, primary_key=True, autoincrement=True)
run_id = Column('run_id', Text, nullable=False)
hostname = Column('hostname', Text, nullable=False)
uid = Column('uid', Text, nullable=False)
block_id = Column('block_id', Text, nullable=False)
cpu_count = Column('cpu_count', Integer, nullable=False)
total_memory = Column('total_memory', BigInteger, nullable=False)
active = Column('active', Boolean, nullable=False)
worker_count = Column('worker_count', Integer, nullable=False)
python_v = Column('python_v', Text, nullable=False)
timestamp = Column('timestamp', DateTime, nullable=False)
last_heartbeat = Column('last_heartbeat', DateTime, nullable=False)
class Block(Base):
__tablename__ = BLOCK
run_id = Column('run_id', Text, nullable=False)
executor_label = Column('executor_label', Text, nullable=False)
block_id = Column('block_id', Text, nullable=False)
job_id = Column('job_id', Text, nullable=True)
timestamp = Column('timestamp', DateTime, nullable=False)
status = Column("status", Text, nullable=False)
__table_args__ = (
PrimaryKeyConstraint('run_id', 'block_id', 'executor_label', 'timestamp'),
)
class Resource(Base):
__tablename__ = RESOURCE
try_id = Column('try_id', Integer, nullable=False)
task_id = Column('task_id', Integer, nullable=False)
run_id = Column('run_id', Text, sa.ForeignKey(
'workflow.run_id'), nullable=False)
timestamp = Column('timestamp', DateTime, nullable=False)
resource_monitoring_interval = Column(
'resource_monitoring_interval', Float, nullable=True)
psutil_process_pid = Column(
'psutil_process_pid', Integer, nullable=True)
psutil_process_memory_percent = Column(
'psutil_process_memory_percent', Float, nullable=True)
psutil_process_children_count = Column(
'psutil_process_children_count', Float, nullable=True)
psutil_process_time_user = Column(
'psutil_process_time_user', Float, nullable=True)
psutil_process_time_system = Column(
'psutil_process_time_system', Float, nullable=True)
psutil_process_memory_virtual = Column(
'psutil_process_memory_virtual', Float, nullable=True)
psutil_process_memory_resident = Column(
'psutil_process_memory_resident', Float, nullable=True)
psutil_process_disk_read = Column(
'psutil_process_disk_read', Float, nullable=True)
psutil_process_disk_write = Column(
'psutil_process_disk_write', Float, nullable=True)
psutil_process_status = Column(
'psutil_process_status', Text, nullable=True)
psutil_cpu_num = Column(
'psutil_cpu_num', Text, nullable=True)
psutil_process_num_ctx_switches_voluntary = Column(
'psutil_process_num_ctx_switches_voluntary', Float, nullable=True)
psutil_process_num_ctx_switches_involuntary = Column(
'psutil_process_num_ctx_switches_involuntary', Float, nullable=True)
__table_args__ = (
PrimaryKeyConstraint('try_id', 'task_id', 'run_id', 'timestamp'),
)
class DatabaseManager:
def __init__(self,
db_url: str = 'sqlite:///runinfo/monitoring.db',
run_dir: str = '.',
logging_level: int = logging.INFO,
batching_interval: float = 1,
batching_threshold: float = 99999,
):
self.workflow_end = False
self.workflow_start_message: Optional[MonitoringMessage] = None
self.run_dir = run_dir
os.makedirs(self.run_dir, exist_ok=True)
logger.propagate = False
set_file_logger(f"{self.run_dir}/database_manager.log", level=logging_level,
format_string="%(asctime)s.%(msecs)03d %(name)s:%(lineno)d [%(levelname)s] [%(threadName)s %(thread)d] %(message)s",
name="database_manager")
logger.debug("Initializing Database Manager process")
self.db = Database(db_url)
self.batching_interval = batching_interval
self.batching_threshold = batching_threshold
self.pending_priority_queue: queue.Queue[TaggedMonitoringMessage] = queue.Queue()
self.pending_node_queue: queue.Queue[MonitoringMessage] = queue.Queue()
self.pending_block_queue: queue.Queue[MonitoringMessage] = queue.Queue()
self.pending_resource_queue: queue.Queue[MonitoringMessage] = queue.Queue()
def start(self,
resource_queue: mpq.Queue) -> None:
self._kill_event = threading.Event()
self._resource_queue_pull_thread = threading.Thread(target=self._migrate_logs_to_internal,
args=(
resource_queue, self._kill_event,),
name="Monitoring-migrate-resource",
daemon=True,
)
self._resource_queue_pull_thread.start()
"""
maintain a set to track the tasks that are already INSERTed into database
to prevent race condition that the first resource message (indicate 'running' state)
arrives before the first task message. In such a case, the resource table
primary key would be violated.
If that happens, the message will be added to deferred_resource_messages and processed later.
"""
inserted_tasks: Set[object] = set()
"""
like inserted_tasks but for task,try tuples
"""
inserted_tries: Set[Any] = set()
# for any task ID, we can defer exactly one message, which is the
# assumed-to-be-unique first message (with first message flag set).
# The code prior to this patch will discard previous message in
# the case of multiple messages to defer.
deferred_resource_messages: MonitoringMessage = {}
exception_happened = False
while (not self._kill_event.is_set() or
self.pending_priority_queue.qsize() != 0 or self.pending_resource_queue.qsize() != 0 or
self.pending_node_queue.qsize() != 0 or self.pending_block_queue.qsize() != 0 or
resource_queue.qsize() != 0):
"""
WORKFLOW_INFO and TASK_INFO messages (i.e. priority messages)
"""
try:
logger.debug("""Checking STOP conditions: {}, {}, {}, {}, {}, {}""".format(
self._kill_event.is_set(),
self.pending_priority_queue.qsize() != 0, self.pending_resource_queue.qsize() != 0,
self.pending_node_queue.qsize() != 0, self.pending_block_queue.qsize() != 0,
resource_queue.qsize() != 0))
# This is the list of resource messages which can be reprocessed as if they
# had just arrived because the corresponding first task message has been
# processed (corresponding by task id)
reprocessable_first_resource_messages = []
# end-of-task-run status messages - handled in similar way as
# for last resource messages to try to have symmetry... this
# needs a type annotation though reprocessable_first_resource_messages
# doesn't... not sure why. Too lazy right now to figure out what,
# if any, more specific type than Any the messages have.
reprocessable_last_resource_messages: List[Any] = []
# Get a batch of priority messages
priority_messages = self._get_messages_in_batch(self.pending_priority_queue)
if priority_messages:
logger.debug(
"Got {} messages from priority queue".format(len(priority_messages)))
task_info_update_messages, task_info_insert_messages, task_info_all_messages = [], [], []
try_update_messages, try_insert_messages, try_all_messages = [], [], []
for msg_type, msg in priority_messages:
if msg_type == MessageType.WORKFLOW_INFO:
if "python_version" in msg: # workflow start message
logger.debug(
"Inserting workflow start info to WORKFLOW table")
self._insert(table=WORKFLOW, messages=[msg])
self.workflow_start_message = msg
else: # workflow end message
logger.debug(
"Updating workflow end info to WORKFLOW table")
self._update(table=WORKFLOW,
columns=['run_id', 'tasks_failed_count',
'tasks_completed_count', 'time_completed'],
messages=[msg])
self.workflow_end = True
elif msg_type == MessageType.TASK_INFO:
task_try_id = str(msg['task_id']) + "." + str(msg['try_id'])
task_info_all_messages.append(msg)
if msg['task_id'] in inserted_tasks:
task_info_update_messages.append(msg)
else:
inserted_tasks.add(msg['task_id'])
task_info_insert_messages.append(msg)
try_all_messages.append(msg)
if task_try_id in inserted_tries:
try_update_messages.append(msg)
else:
inserted_tries.add(task_try_id)
try_insert_messages.append(msg)
# check if there is a left_message for this task
if task_try_id in deferred_resource_messages:
reprocessable_first_resource_messages.append(
deferred_resource_messages.pop(task_try_id))
else:
raise RuntimeError("Unexpected message type {} received on priority queue".format(msg_type))
logger.debug("Updating and inserting TASK_INFO to all tables")
logger.debug("Updating {} TASK_INFO into workflow table".format(len(task_info_update_messages)))
self._update(table=WORKFLOW,
columns=['run_id', 'tasks_failed_count',
'tasks_completed_count'],
messages=task_info_all_messages)
if task_info_insert_messages:
self._insert(table=TASK, messages=task_info_insert_messages)
logger.debug(
"There are {} inserted task records".format(len(inserted_tasks)))
if task_info_update_messages:
logger.debug("Updating {} TASK_INFO into task table".format(len(task_info_update_messages)))
self._update(table=TASK,
columns=['task_time_invoked',
'task_time_returned',
'run_id', 'task_id',
'task_fail_count',
'task_fail_cost',
'task_hashsum',
'task_inputs'],
messages=task_info_update_messages)
logger.debug("Inserting {} task_info_all_messages into status table".format(len(task_info_all_messages)))
self._insert(table=STATUS, messages=task_info_all_messages)
if try_insert_messages:
logger.debug("Inserting {} TASK_INFO to try table".format(len(try_insert_messages)))
self._insert(table=TRY, messages=try_insert_messages)
logger.debug(
"There are {} inserted task records".format(len(inserted_tasks)))
if try_update_messages:
logger.debug("Updating {} TASK_INFO into try table".format(len(try_update_messages)))
self._update(table=TRY,
columns=['run_id', 'task_id', 'try_id',
'task_fail_history',
'task_try_time_launched',
'task_try_time_returned',
'task_joins'],
messages=try_update_messages)
"""
NODE_INFO messages
"""
node_info_messages = self._get_messages_in_batch(self.pending_node_queue)
if node_info_messages:
logger.debug(
"Got {} messages from node queue".format(len(node_info_messages)))
self._insert(table=NODE, messages=node_info_messages)
"""
BLOCK_INFO messages
"""
block_info_messages = self._get_messages_in_batch(self.pending_block_queue)
if block_info_messages:
logger.debug(
"Got {} messages from block queue".format(len(block_info_messages)))
# block_info_messages is possibly a nested list of dict (at different polling times)
# Each dict refers to the info of a job/block at one polling time
block_messages_to_insert: List[Any] = []
for block_msg in block_info_messages:
block_messages_to_insert.extend(block_msg)
self._insert(table=BLOCK, messages=block_messages_to_insert)
"""
Resource info messages
"""
resource_messages = self._get_messages_in_batch(self.pending_resource_queue)
if resource_messages:
logger.debug(
"Got {} messages from resource queue, "
"{} reprocessable as first messages, "
"{} reprocessable as last messages".format(len(resource_messages),
len(reprocessable_first_resource_messages),
len(reprocessable_last_resource_messages)))
insert_resource_messages = []
for msg in resource_messages:
task_try_id = str(msg['task_id']) + "." + str(msg['try_id'])
if msg['first_msg']:
# Update the running time to try table if first message
msg['task_status_name'] = States.running.name
msg['task_try_time_running'] = msg['timestamp']
if task_try_id in inserted_tries: # TODO: needs to become task_id and try_id, and check against inserted_tries
reprocessable_first_resource_messages.append(msg)
else:
if task_try_id in deferred_resource_messages:
logger.error(
"Task {} already has a deferred resource message. "
"Discarding previous message.".format(msg['task_id'])
)
deferred_resource_messages[task_try_id] = msg
elif msg['last_msg']:
# This assumes that the primary key has been added
# to the try table already, so doesn't use the same
# deferral logic as the first_msg case.
msg['task_status_name'] = States.running_ended.name
reprocessable_last_resource_messages.append(msg)
else:
# Insert to resource table if not first/last (start/stop) message message
insert_resource_messages.append(msg)
if insert_resource_messages:
self._insert(table=RESOURCE, messages=insert_resource_messages)
if reprocessable_first_resource_messages:
self._insert(table=STATUS, messages=reprocessable_first_resource_messages)
self._update(table=TRY,
columns=['task_try_time_running',
'run_id', 'task_id', 'try_id',
'block_id', 'hostname'],
messages=reprocessable_first_resource_messages)
if reprocessable_last_resource_messages:
self._insert(table=STATUS, messages=reprocessable_last_resource_messages)
except Exception:
logger.exception(
"Exception in db loop: this might have been a malformed message, "
"or some other error. monitoring data may have been lost"
)
exception_happened = True
if exception_happened:
raise RuntimeError("An exception happened sometime during database processing and should have been logged in database_manager.log")
@wrap_with_logs(target="database_manager")
def _migrate_logs_to_internal(self, logs_queue: queue.Queue, kill_event: threading.Event) -> None:
logger.info("Starting _migrate_logs_to_internal")
while not kill_event.is_set() or logs_queue.qsize() != 0:
logger.debug("Checking STOP conditions: kill event: %s, queue has entries: %s",
kill_event.is_set(), logs_queue.qsize() != 0)
try:
x = logs_queue.get(timeout=0.1)
except queue.Empty:
continue
else:
if x == 'STOP':
self.close()
else:
self._dispatch_to_internal(x)
def _dispatch_to_internal(self, x: Tuple) -> None:
assert isinstance(x, tuple)
assert len(x) == 2, "expected message tuple to have exactly two elements"
if x[0] in [MessageType.WORKFLOW_INFO, MessageType.TASK_INFO]:
self.pending_priority_queue.put(cast(Any, x))
elif x[0] == MessageType.RESOURCE_INFO:
body = x[1]
self.pending_resource_queue.put(body)
elif x[0] == MessageType.NODE_INFO:
assert len(x) == 2, "expected NODE_INFO tuple to have exactly two elements"
logger.info("Will put {} to pending node queue".format(x[1]))
self.pending_node_queue.put(x[1])
elif x[0] == MessageType.BLOCK_INFO:
logger.info("Will put {} to pending block queue".format(x[1]))
self.pending_block_queue.put(x[-1])
else:
logger.error("Discarding message of unknown type {}".format(x[0]))
def _update(self, table: str, columns: List[str], messages: List[MonitoringMessage]) -> None:
try:
done = False
while not done:
try:
self.db.update(table=table, columns=columns, messages=messages)
done = True
except sa.exc.OperationalError as e:
# This code assumes that an OperationalError is something that will go away eventually
# if retried - for example, the database being locked because someone else is readying
# the tables we are trying to write to. If that assumption is wrong, then this loop
# may go on forever.
logger.warning("Got a database OperationalError. "
"Ignoring and retrying on the assumption that it is recoverable: {}".format(e))
self.db.rollback()
time.sleep(1) # hard coded 1s wait - this should be configurable or exponential backoff or something
except KeyboardInterrupt:
logger.exception("KeyboardInterrupt when trying to update Table {}".format(table))
try:
self.db.rollback()
except Exception:
logger.exception("Rollback failed")
raise
except Exception:
logger.exception("Got exception when trying to update table {}".format(table))
try:
self.db.rollback()
except Exception:
logger.exception("Rollback failed")
def _insert(self, table: str, messages: List[MonitoringMessage]) -> None:
try:
done = False
while not done:
try:
self.db.insert(table=table, messages=messages)
done = True
except sa.exc.OperationalError as e:
# hoping that this is a database locked error during _update, not some other problem
logger.warning("Got a database OperationalError. "
"Ignoring and retrying on the assumption that it is recoverable: {}".format(e))
self.db.rollback()
time.sleep(1) # hard coded 1s wait - this should be configurable or exponential backoff or something
except KeyboardInterrupt:
logger.exception("KeyboardInterrupt when trying to update Table {}".format(table))
try:
self.db.rollback()
except Exception:
logger.exception("Rollback failed")
raise
except Exception:
logger.exception("Got exception when trying to insert to table {}".format(table))
try:
self.db.rollback()
except Exception:
logger.exception("Rollback failed")
def _get_messages_in_batch(self, msg_queue: "queue.Queue[X]") -> List[X]:
messages: List[X] = []
start = time.time()
while True:
if time.time() - start >= self.batching_interval or len(messages) >= self.batching_threshold:
break
try:
x = msg_queue.get(timeout=0.1)
# logger.debug("Database manager receives a message {}".format(x))
except queue.Empty:
logger.debug("Database manager has not received any message.")
break
else:
messages.append(x)
return messages
def close(self) -> None:
logger.info("Database Manager cleanup initiated.")
if not self.workflow_end and self.workflow_start_message:
logger.info("Logging workflow end info to database due to abnormal exit")
time_completed = datetime.datetime.now()
msg = {'time_completed': time_completed,
'workflow_duration': (time_completed - self.workflow_start_message['time_began']).total_seconds()}
self.workflow_start_message.update(msg)
self._update(table=WORKFLOW,
columns=['run_id', 'time_completed',
'workflow_duration'],
messages=[self.workflow_start_message])
self.batching_interval = float('inf')
self.batching_threshold = float('inf')
self._kill_event.set()
@wrap_with_logs(target="database_manager")
@typeguard.typechecked
def dbm_starter(exception_q: mpq.Queue,
resource_msgs: mpq.Queue,
db_url: str,
run_dir: str,
logging_level: int) -> None:
"""Start the database manager process
The DFK should start this function. The args, kwargs match that of the monitoring config
"""
setproctitle("parsl: monitoring database")
try:
dbm = DatabaseManager(db_url=db_url,
run_dir=run_dir,
logging_level=logging_level)
logger.info("Starting dbm in dbm starter")
dbm.start(resource_msgs)
except KeyboardInterrupt:
logger.exception("KeyboardInterrupt signal caught")
dbm.close()
raise
except Exception as e:
logger.exception("dbm.start exception")
exception_q.put(("DBM", str(e)))
dbm.close()
logger.info("End of dbm_starter")
|