manager.py 21.1 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

import time
import socket
import os
import six
19
import copy
20 21
import logging
import signal
K
kuizhiqing 已提交
22
import random
23 24
import threading
import traceback
25
import subprocess
26 27
from paddle.distributed.fleet import cloud_utils
from paddle.distributed.fleet import launch_utils
28 29

logger = logging.getLogger("ELASTIC")
30 31 32 33 34 35
logger.setLevel(logging.INFO)
formatter = logging.Formatter(
    fmt='%(name)s %(levelname)s %(asctime)s %(message)s')
ch = logging.StreamHandler()
ch.setFormatter(formatter)
logger.addHandler(ch)
36 37 38

ELASTIC_EXIT_CODE = 101

39 40 41 42 43 44 45 46 47 48 49 50
# wait for timeout, unit: seconds
ELASTIC_TIMEOUT = 2 * 60

# keepalived ttl, unit: seconds
ELASTIC_TTL = 60


# 1: Fault tolerance, 2: Elastic
class ElasticLevel:
    FAULT_TOLERANCE = 1
    ELASTIC = 2

51 52 53 54 55 56 57 58 59 60 61 62 63 64 65

class ElasticStatus:
    COMPLETED = "completed"
    ERROR = "error"
    HOLD = "hold"
    RESTART = "restart"
    EXIT = "exit"


class LauncherInterface(object):
    def __init__(self, args):
        self.args = args
        self.procs = []

    def _terminate_procs(self):
K
kuizhiqing 已提交
66
        # try to terminate process by group, this happend in multiprocess senario in user process
K
kuizhiqing 已提交
67 68 69 70 71 72 73 74
        if os.name != 'nt':
            for p in self.procs:
                if p.proc.poll() is None:
                    os.killpg(os.getpgid(p.proc.pid), signal.SIGTERM)
                    if p.log_fn:
                        p.log_fn.close()
                    logger.info("terminate process group gid:{}".format(
                        p.proc.pid))
K
kuizhiqing 已提交
75

K
kuizhiqing 已提交
76
            time.sleep(1)
77 78 79 80 81 82 83 84 85 86 87 88 89 90 91
        for p in self.procs:
            if p.proc.poll() is None:
                p.proc.terminate()
                if p.log_fn:
                    p.log_fn.close()
                logger.info("terminate process id:{}".format(p.proc.pid))

        for step in range(0, 50):
            alive = False
            for p in self.procs:
                if p.proc.poll() is None:  # not termniate
                    os.kill(p.proc.pid, signal.SIGKILL)
                    alive = True

            if not alive:
K
kuizhiqing 已提交
92
                logger.info("terminated all the procs")
93 94 95 96 97 98 99 100 101 102 103 104 105
                return True

            time.sleep(1)
        return False

    def _check_procs(self):
        alive = False
        result = None
        for p in self.procs:
            ret = p.proc.poll()
            if ret is None:
                alive = True
            elif ret != 0:
K
kuizhiqing 已提交
106 107 108 109
                logger.error("ABORT!!! ABORT!!! ABORT!!!")
                logger.error(
                    "ERROR rank {} error with exit code {}, check log for detail.".
                    format(p.rank, ret))
110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126
                result = ret
        if not alive and result is None:
            return 0
        else:
            return result

    def launch(self):
        raise NotImplementedError

    def stop(self):
        raise NotImplementedError

    def watch(self):
        raise NotImplementedError


class ElasticManager(object):
127
    def __init__(self, args, etcd_client):
128 129 130 131

        self.args = args
        server = args.elastic_server or os.getenv('PADDLE_ELASTIC_SERVER')
        name = args.job_id or os.getenv('PADDLE_ELASTIC_JOB_ID')
132
        self.min_np, self.max_np = self._parse_np(args.np)
133 134 135 136
        host = args.host or os.getenv('POD_IP')
        scale = args.scale or int(os.getenv('PADDLE_ELASTIC_SCALE', 0))
        force = args.force or os.getenv('PADDLE_ELASTIC_FORCE')

137
        self.host = host if host else self._get_host()
138

139 140 141 142 143 144 145
        (self.device_mode,
         self.devices_per_proc) = launch_utils.get_device_proc_info(args)

        self.elastic_timeout = int(
            os.getenv('PADDLE_ELASTIC_TIMEOUT', ELASTIC_TIMEOUT))
        elastic_ttl = int(os.getenv('PADDLE_ELASTIC_TTL', ELASTIC_TTL))

146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166
        self.start_port = None
        if cloud_utils.use_paddlecloud():
            self.trainers = os.getenv('PADDLE_TRAINERS', '')
            self.np = len(self.trainers.split(","))
            self.start_port = int(os.getenv("PADDLE_PORT", "6170"))
            self.dist_endpoints = os.getenv('DISTRIBUTED_TRAINER_ENDPOINTS', '')
            trainer_endpoints = os.getenv('PADDLE_TRAINER_ENDPOINTS', '')
            self.trainer_endpoints_list = trainer_endpoints.split(",")
        else:
            self.trainers = args.ips or os.getenv('PADDLE_TRAINERS', '')
            node_ips = self.trainers.split(",")
            self.np = len(node_ips)
            self.start_port = int(os.getenv("FLAGS_START_PORT", "6170"))
            self.dist_endpoints = self._host_to_endpoints(
                node_ips, self.devices_per_proc, self.start_port)
            self.trainer_endpoints_list = [
                "%s:%d" % (ip, self.start_port) for ip in node_ips
            ]

        self.curr_host = "%s:%d" % (self.host, self.start_port)
        logger.info(f'start job with np={self.np}')
167
        logger.info(
168
            f"trainers={self.trainers}, trainer_endpoints_list={self.trainer_endpoints_list}"
169 170 171 172
        )

        # auto correct the value of elastic_level
        # 1: Fault tolerant, 2: Elastic
173
        self.elastic_level = int(
174 175 176 177 178 179 180 181 182
            os.getenv('PADDLE_ELASTIC_FAULT_TOLERANC_LEVEL',
                      ElasticLevel.FAULT_TOLERANCE))
        if self.min_np == self.max_np or \
                (self.min_np > 0 and self.max_np == 0):
            self.elastic_level = ElasticLevel.FAULT_TOLERANCE
            logger.info(f'start job with ElasticLevel.FAULT_TOLERANCE')
        if self.min_np > 0 and self.max_np > self.min_np:
            self.elastic_level = ElasticLevel.ELASTIC
            logger.info(f'start job with ElasticLevel.ELASTIC')
183

K
kuizhiqing 已提交
184 185 186 187 188 189 190 191
        # compatible with kuberntes service discovery
        if not server and os.getenv(
                'PADDLE_ELASTIC_ETCD_SERVICE_HOST') and os.getenv(
                    'PADDLE_ELASTIC_ETCD_SERVICE_PORT'):
            server = '{}:{}'.format(
                os.getenv('PADDLE_ELASTIC_ETCD_SERVICE_HOST'),
                os.getenv('PADDLE_ELASTIC_ETCD_SERVICE_PORT'))

192 193 194 195 196 197
        logger.debug('init with server {} host {}'.format(server, host))

        self.hosts = []
        self.stopped = False

        self.sigint = 0
K
kuizhiqing 已提交
198
        self.need_sync = False
199

200 201 202
        self.elastic_startup_time = None

        if not server or ':' not in server or not name or not self.np:
203 204
            logger.info(
                'Elastic is not enabled with server {} name {} and np {}'.
205
                format(server, name, self.np))
206 207 208 209 210
            self.enable = False
            return
        else:
            self.enable = True

211
        self.etcd = etcd_client
212 213 214

        # etcd data
        self.prefix = "/paddle/" + name
K
kuizhiqing 已提交
215
        self.node_prefix = self.prefix + '/nodes'
216 217
        self.np_path = self.prefix + '/np'
        self.endpoints_path = self.prefix + '/endpoints'
K
kuizhiqing 已提交
218 219 220 221 222

        node_tag = ''.join(
            random.choice('abcdefghijklmnopqrstuvwxyz') for _ in range(6))
        self.host_path = '{}/{}{}'.format(self.node_prefix, node_tag,
                                          time.time())
223 224 225 226 227 228
        '''
        0 group mode, be aware of healthy status of other workers
        1 decouple mode, check own status only
        '''
        self.etcd.put(self.prefix, b'0')

229
        # register callback
230
        def host_call_back(event):
231 232 233 234 235
            self.hosts = [
                six.ensure_str(i[0])
                for i in self.etcd.get_prefix(self.node_prefix)
            ]
            logger.info(
236
                f"host_call_back curr_host={self.curr_host}, hosts:{self.hosts}")
237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254
            self.need_sync = True
            self.elastic_startup_time = None

        host_watch = self.etcd.add_watch_prefix_callback(self.node_prefix,
                                                         host_call_back)
        host_lease = self.etcd.lease(elastic_ttl)

        # register etcd lease heartbeat
        def lease_heartbeat():
            while True:
                try:
                    host_lease.refresh()

                    hosts = [
                        six.ensure_str(i[0])
                        for i in self.etcd.get_prefix(self.node_prefix)
                    ]
                    logger.info(
255
                        f"[lease_heartbeat] curr_host={self.curr_host}, hosts={hosts}"
256
                    )
257
                    if self.curr_host not in hosts:
258
                        logger.info(
259
                            f"[lease_heartbeat] register host={self.curr_host}")
260
                        self.etcd.put(self.host_path,
261
                                      six.b(self.curr_host),
262 263 264 265 266 267 268 269 270 271 272
                                      lease=host_lease)
                except Exception as e:
                    logger.error("[lease_heartbeat] internal error:{} {}".
                                 format(e, traceback.format_exc()))
                    break
                time.sleep(elastic_ttl / 3)

        keepalived_thread = threading.Thread(
            name='lease_heartbeat', target=lease_heartbeat, daemon=True)
        keepalived_thread.start()

273
        self.etcd.put(self.host_path, six.b(self.curr_host), lease=host_lease)
274 275 276

        # endpoints handle DISTRIBUTED_TRAINER_ENDPOINTS and PADDLE_TRAINERS
        self.etcd.put(self.endpoints_path,
277
                      six.b('{}|{}'.format(self.dist_endpoints, self.trainers)))
278 279

        def endpoints_call_back(event):
280
            if not self.dist_endpoints:
281 282
                return
            edps = six.ensure_str(self.etcd.get(self.endpoints_path)[0] or '')
283
            self.dist_endpoints, self.trainers = edps.split('|')
284
            logger.info("set DISTRIBUTED_TRAINER_ENDPOINTS {} ".format(
285
                self.dist_endpoints))
286 287 288 289 290
            logger.info("set PADDLE_TRAINERS {} ".format(self.trainers))

        endpoints_watch = self.etcd.add_watch_callback(self.endpoints_path,
                                                       endpoints_call_back)

291
        self.watches = [host_watch, endpoints_watch]
K
kuizhiqing 已提交
292 293
        self.launcher = None

294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313
    def _host_to_endpoints(self,
                           ip_port_list: list,
                           devices_per_proc: list,
                           start_port: int=6170) -> str:
        endpoint_list = []
        for ip_port in ip_port_list:
            endpoints = ip_port.split(":")
            if len(endpoints) == 2:
                ip = endpoints[0]
                port = int(endpoints[1])
            else:
                ip = endpoints
                port = start_port

            ports = [x for x in range(port, port + len(devices_per_proc))]
            endpoint_list.extend(["%s:%d" % (ip, port) for port in ports])

        dist_endpoints = ','.join(endpoint_list)
        return dist_endpoints

314 315 316
    def exit(self, completed=False):
        logger.info('manager exist completed {}'.format(completed))

K
kuizhiqing 已提交
317 318
        if self.launcher:
            self.launcher.stop()
K
kuizhiqing 已提交
319

320 321 322 323 324 325 326 327 328 329 330 331 332 333
        if not self.enable:
            return

        if completed:
            self.etcd.put(self.prefix, b'1')

        for watch in self.watches:
            self.etcd.cancel_watch(watch)
        self.etcd.delete(self.host_path)

        hosts = [i for i in self.etcd.get_prefix(self.node_prefix)]
        if len(hosts) == 0:
            self.etcd.delete_prefix(self.prefix)

334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349
    def pre_hook(self):
        if not self.args.elastic_pre_hook:
            logger.info("skip pre_hook")
            return
        current_env = copy.copy(os.environ.copy())
        out, err = subprocess.Popen(
            self.args.elastic_pre_hook,
            env=current_env,
            stdout=subprocess.PIPE,
            stderr=subprocess.PIPE,
            shell=True).communicate()
        if err:
            logger.warn("pre_hook exec failed")
        else:
            logger.info(f"pre_hook exec result: {out.decode('utf-8').strip()}")

350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373
    def _parse_np(self, np: str):
        """
        np format is "MIN" or "MIN:MAX" 
        """
        np_str = np or os.getenv('PADDLE_ELASTIC_NP', "0")
        np_dict = np_str.split(":")
        min_np = max_np = 0
        if len(np_dict) == 1:
            # Fault tolerant
            min_np = int(np_dict[0])
            min_np = 1 if min_np <= 0 else min_np
            max_np = 1
        elif len(np_dict) == 2:
            # Elastic
            min_np = int(np_dict[0])
            max_np = int(np_dict[1])
            min_np = 1 if min_np <= 0 else min_np
            max_np = min_np if min_np > max_np else max_np
        else:
            raise ValueError(
                f'the np={np} needs to be in "MIN" or "MIN:MAX" format')

        return min_np, max_np

374 375 376 377 378 379 380 381 382 383 384 385
    def _get_host(self):
        try:
            return socket.gethostbyname(socket.getfqdn(socket.gethostname()))
        except:
            return '127.0.0.1'

    def _completed(self):
        if not self.enable:
            return True

        return int(self.etcd.get(self.prefix)[0]) == 1

386 387 388
    def _match(self, host_list: list=None):
        if host_list:
            self.hosts = host_list
389
        else:
390 391 392 393
            self.hosts = [
                six.ensure_str(i[0])
                for i in self.etcd.get_prefix(self.node_prefix)
            ]
394

395 396 397 398 399
        if self.elastic_level == ElasticLevel.FAULT_TOLERANCE:
            if len(self.hosts) == self.np:
                return True
            else:
                return False
400

401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423
        if self.elastic_level == ElasticLevel.ELASTIC:
            hosts_num = len(self.hosts)
            if hosts_num == self.np:
                return True

            if not self.elastic_startup_time:
                self.elastic_startup_time = time.time()
            if hosts_num == self.max_np:
                self.elastic_startup_time = None
                return True
            elif hosts_num >= self.min_np and hosts_num < self.max_np:
                interval_time = time.time() - self.elastic_startup_time
                if interval_time <= self.elastic_timeout:
                    logger.info(
                        f"wait for timeout, you can set value by PADDLE_ELASTIC_TIMEOUT, \
                        hosts_num={hosts_num}, min_np={self.min_np}, \
                        interval_time={interval_time}, elastic_timeout={self.elastic_timeout}"
                    )
                    return False
                return True
            else:
                self.elastic_startup_time = None
                return False
424

425 426 427 428 429 430
        return False

    def _update_endpoint(self, endpoints, hosts):
        self.etcd.put(self.endpoints_path,
                      six.b('{}|{}'.format(endpoints, hosts)))

431
    def _update_fault_tolrance(self):
432
        rank = int(os.getenv('PADDLE_TRAINER_ID', -1))
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
        if self.curr_host in self.dist_endpoints:
            os.environ['DISTRIBUTED_TRAINER_ENDPOINTS'] = self.dist_endpoints
            os.environ['PADDLE_TRAINERS'] = self.trainers
            logger.info("update env DISTRIBUTED_TRAINER_ENDPOINTS {} ".format(
                self.dist_endpoints))
            logger.info("update env PADDLE_TRAINERS {} ".format(self.trainers))
            return

        # fault tolerance 
        idx = self.hosts.index(self.curr_host)

        # swap if self.host not in the right position
        if rank >= 0:
            self.hosts[idx] = self.hosts[rank]
            self.hosts[rank] = self.curr_host
        else:
            os.environ['PADDLE_TRAINER_ID'] = '{}'.format(idx)
        hosts = ','.join([host_port.split(":")[0] for host_port in self.hosts])
        self.args.ips = hosts
        os.environ['PADDLE_TRAINERS'] = hosts

    def _update_elastic_scale_out(self):
        host_endpoints = copy.deepcopy(self.trainer_endpoints_list)
        logger.info(
            f"elastic scale out, from {len(self.hosts)} to {self.np}, hosts={self.hosts}, host_endpoints={host_endpoints}"
        )
459

460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479
        for curr_host_port in self.hosts:
            if curr_host_port not in host_endpoints:
                host_endpoints.append(curr_host_port)

        os.environ['PADDLE_TRAINER_ID'] = '{}'.format(
            host_endpoints.index(self.curr_host))
        hosts = ','.join(
            [host_port.split(":")[0] for host_port in host_endpoints])
        self.args.ips = hosts
        os.environ['PADDLE_TRAINERS'] = hosts
        self.np = len(host_endpoints)
        os.environ['PADDLE_TRAINER_ENDPOINTS'] = ','.join(host_endpoints)
        os.environ['DISTRIBUTED_TRAINER_ENDPOINTS'] = self.dist_endpoints
        self.trainer_endpoints_list = host_endpoints

    def _update_elastic_scale_in(self):
        host_endpoints = copy.deepcopy(self.trainer_endpoints_list)
        logger.info(
            f"elastic scale in, from {self.np} to {len(self.hosts)}, hosts={self.hosts}, host_endpoints={host_endpoints}"
        )
480

481 482 483 484 485 486 487 488 489 490 491 492
        # If scale in node from the first of the rank list, you need to minimize the movement of the rank
        # eg: 
        #   the source trainers is:10.10.10.0,10.10.10.1,10.10.10.2,10.10.10.3
        #   10.10.10.0 is removed
        #   the new trainers is:10.10.10.3,10.10.10.1,10.10.10.2
        #   In this case, the rank of 10.10.10.1 and 10.10.10.2 remains unchanged, while the rank of 10.10.10.3 is set to rank0
        endpoints_dict = dict()
        unsorted_endpoints = []
        for id, host_port in enumerate(self.hosts):
            idx = host_endpoints.index(host_port)
            if idx <= len(self.hosts) - 1 and not endpoints_dict.get(idx):
                endpoints_dict[idx] = host_port
493
            else:
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
                unsorted_endpoints.append(host_port)

        idle_index = 0
        sorted_endpoints = []
        for idx in range(len(self.hosts)):
            if not endpoints_dict.get(idx) and len(unsorted_endpoints) > 0:
                endpoints_dict[idx] = unsorted_endpoints[idle_index]
                idle_index += 1

            sorted_endpoints.append(endpoints_dict.get(idx))

        logger.info(f"elastic scale in, sorted_endpoints={sorted_endpoints}")
        self.trainer_endpoints_list = sorted_endpoints

        ip_list = [ip_port.split(":")[0] for ip_port in sorted_endpoints]
        hosts = ','.join(ip_list)
        new_endpoints = self._host_to_endpoints(sorted_endpoints,
                                                self.devices_per_proc)

        self.args.ips = hosts
        os.environ['PADDLE_TRAINER_ID'] = '{}'.format(
            sorted_endpoints.index(self.curr_host))
        os.environ['PADDLE_TRAINERS'] = hosts
        self.np = len(sorted_endpoints)
        os.environ['PADDLE_TRAINER_ENDPOINTS'] = ','.join(sorted_endpoints)
        os.environ['DISTRIBUTED_TRAINER_ENDPOINTS'] = new_endpoints
        self._update_endpoint(new_endpoints, hosts)

    def _update_hosts(self):
        assert len(self.hosts) != 0, 'hosts empty'
        if self.elastic_level == ElasticLevel.FAULT_TOLERANCE:
            self._update_fault_tolrance()
526
        else:
527 528 529 530 531 532 533 534
            # elastic
            if len(self.hosts) == self.np:
                logger.info(f"elastic startup, hosts={self.hosts}")
                self._update_fault_tolrance()

            elif len(self.hosts) > self.np:
                # scale out
                self._update_elastic_scale_out()
535
            else:
536 537
                # scale in
                self._update_elastic_scale_in()
538 539 540 541 542

    def wait(self):
        if not self.enable:
            return

K
kuizhiqing 已提交
543
        idx = 1
544 545 546 547 548 549 550
        while not self.stopped:
            if self._match():
                logger.info('ready with hosts {}'.format(self.hosts))
                self._update_hosts()
                return
            logger.info('not ready for np {} with hosts {}'.format(self.np,
                                                                   self.hosts))
K
kuizhiqing 已提交
551
            idx += 1
K
kuizhiqing 已提交
552 553
            time.sleep(2)

554 555 556 557 558 559 560 561 562 563 564
        return

    def run(self, launcher):
        if self.stopped:
            return

        self.launcher = launcher(self.args)
        self.launcher.launch()

    def watch(self):

K
kuizhiqing 已提交
565 566 567
        if self.need_sync:
            self.need_sync = False

568 569
        while not self.stopped:
            ret = self.launcher.watch()
570
            logger.debug(f"launcher.watch():{ret}")
571 572 573 574 575 576 577 578

            if ret is not None:  # self terminated
                logger.info('job exit with code {}'.format(ret))
                # process is completed if ret >= 0 or error else
                completed = True if ret == 0 else False
                self.exit(completed=completed)
                if completed:
                    return ElasticStatus.COMPLETED
579
                if self.elastic_level == ElasticLevel.FAULT_TOLERANCE:
580 581 582 583
                    return ElasticStatus.RESTART
                else:
                    return ElasticStatus.ERROR

K
kuizhiqing 已提交
584
            if not self._completed() and (not self._match() or self.need_sync):
585 586 587
                self.launcher.stop()
                return ElasticStatus.HOLD

K
kuizhiqing 已提交
588
            time.sleep(2)
589

K
kuizhiqing 已提交
590 591
        if self.launcher:
            self.launcher.stop()
592

593 594 595 596 597 598 599
        return ElasticStatus.EXIT

    def signal_handler(self, sigint, frame):
        if self.enable:
            self.exit()
        self.sigint = sigint
        self.stopped = True