driver.py

# Copyright 2020 Uber Technologies, Inc. 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 logging
import os
import queue
import threading
import time
import warnings

from collections import defaultdict

from horovod.runner.common.util import hosts, timeout
from horovod.runner.elastic.discovery import HostManager
from horovod.runner.elastic.registration import WorkerStateRegistry
from horovod.runner.elastic.worker import HostUpdateResult, WorkerNotificationClient


DISCOVER_HOSTS_FREQUENCY_SECS = 1.0
ELASTIC_TIMEOUT_SECS = 600


def _epoch_time_s():
    return int(time.time())


class Results(object):
    def __init__(self, error_message, worker_results):
        self.error_message = error_message
        self.worker_results = worker_results


class ResultsRecorder(object):
    def __init__(self):
        self._error_message = None
        self._worker_results = {}
        self._worker_threads = queue.Queue()

    def expect(self, worker_thread):
        self._worker_threads.put(worker_thread)

    def set_error_message(self, error_message):
        self._error_message = error_message

    def add_result(self, key, value):
        if key in self._worker_results:
            return
        self._worker_results[key] = value

    def get_results(self):
        while not self._worker_threads.empty():
            worker_thread = self._worker_threads.get()
            worker_thread.join()
        return Results(self._error_message, self._worker_results)


class ElasticDriver(object):
    def __init__(self, rendezvous,
                 discovery,
                 min_num_proc,
                 max_num_proc,
                 timeout=None,
                 reset_limit=None,
                 cooldown_range=None,
                 verbose=0):
        self._rendezvous = rendezvous
        self._host_manager = HostManager(discovery, cooldown_range)
        self._min_num_proc = min_num_proc
        self._max_num_proc = max_num_proc
        self._verbose = verbose

        self._host_assignments = {}
        self._rank_assignments = {}
        self._world_size = 0

        self._wait_hosts_cond = threading.Condition()
        self._timeout = timeout or int(os.getenv('HOROVOD_ELASTIC_TIMEOUT', ELASTIC_TIMEOUT_SECS))

        self._create_worker_fn = None
        self._worker_clients = {}

        self._worker_registry = WorkerStateRegistry(self, self._host_manager, reset_limit=reset_limit)
        self._results = ResultsRecorder()
        self._shutdown = threading.Event()

        self._discovery_thread = threading.Thread(target=self._discover_hosts)
        self._discovery_thread.daemon = True
        self._discovery_thread.start()

    def start(self, num_proc, create_worker_fn):
        self._create_worker_fn = create_worker_fn
        self._activate_workers(num_proc)

    def resume(self):
        self._activate_workers(self._min_num_proc)

    def stop(self, error_message=None):
        self._results.set_error_message(error_message)
        self._shutdown.set()
        self._rendezvous.stop()
        self._discovery_thread.join()

    def finished(self):
        return self._shutdown.is_set()

    def get_results(self):
        return self._results.get_results()

    def register_worker_server(self, host, slot, addresses, secret_key):
        self._worker_clients[(host, slot)] = WorkerNotificationClient(
            addresses, secret_key, self._verbose)

    def get_worker_client(self, slot_info):
        return self._worker_clients.get((slot_info.hostname, slot_info.local_rank))

    def record_ready(self, host, slot):
        self._worker_registry.record_ready(host, slot)

    def world_size(self):
        return self._world_size

    def local_size(self, host):
        return len(self._host_assignments[host])

    def get_slot_info(self, host, slot):
        return self._host_assignments[host][slot] if self.has_rank_assignment(host, slot) \
            else hosts.INVALID_SLOT_INFO

    def get_coordinator_info(self):
        return self._rank_assignments.get(0)

    def has_rank_assignment(self, host, slot):
        if self._host_manager.is_blacklisted(host):
            return False
        return host in self._host_assignments and len(self._host_assignments[host]) > slot

    @property
    def host_assignments(self):
        return self._host_assignments

    def wait_for_available_slots(self, min_num_proc, min_hosts=1):
        extra_message = ' An elastic job also requires that at least two hosts ' \
                        'are available to resolve compatible network interfaces. If you know which interfaces ' \
                        'are compatible in your network, set `--network-interface` to skip this check.' \
            if min_hosts > 1 else ''

        tmout = timeout.Timeout(
            self._timeout,
            message='Timed out waiting for {{activity}}. Please check that you have '
                    f'enough resources to run at least {min_num_proc} Horovod processes.{extra_message}')

        self._wait_hosts_cond.acquire()
        try:
            while True:
                current_hosts = self._host_manager.current_hosts
                avail_slots = current_hosts.count_available_slots()
                logging.debug(f"current available slots: {avail_slots}")
                avail_hosts = len(current_hosts.available_hosts)
                logging.debug(f"current available hosts: {avail_hosts}.")
                if avail_slots >= min_num_proc and avail_hosts >= min_hosts:
                    return current_hosts
                if self._shutdown.is_set():
                    raise RuntimeError('Job has been shutdown, see above error messages for details.')
                self._wait_hosts_cond.wait(tmout.remaining())
                tmout.check_time_out_for('minimum number of slots to become available')
        finally:
            self._wait_hosts_cond.release()

    def _activate_workers(self, min_num_proc):
        logging.info('wait for available slots: {}'.format(min_num_proc))
        current_hosts = self.wait_for_available_slots(min_num_proc)
        pending_slots = self._update_host_assignments(current_hosts)
        self._worker_registry.reset(self.world_size())
        self._start_worker_processes(pending_slots)

    def _discover_hosts(self):
        first_update = True
        while not self._shutdown.is_set():
            self._wait_hosts_cond.acquire()
            try:
                update_res = self._host_manager.update_available_hosts()
                if update_res != HostUpdateResult.no_update:
                    self._notify_workers_host_changes(self._host_manager.current_hosts, update_res)
                    self._wait_hosts_cond.notify_all()
            except RuntimeError as e:
                if first_update:
                    # Misconfiguration, fail the job immediately
                    self._shutdown.set()
                    self._wait_hosts_cond.notify_all()
                    raise
                # Transient error, retry until timeout
                logging.warning(str(e))
            finally:
                self._wait_hosts_cond.release()
            first_update = False
            self._shutdown.wait(DISCOVER_HOSTS_FREQUENCY_SECS)

    def _notify_workers_host_changes(self, current_hosts, update_res):
        next_host_assignments = {}
        if current_hosts.count_available_slots() >= self._min_num_proc:
            # Assignments are required to be stable via contract
            next_host_assignments, _ = self._get_host_assignments(current_hosts)

        if next_host_assignments == self.host_assignments:
            # Skip notifying workers when host changes would not result in changes of host assignments
            logging.debug('no host assignment changes, skipping notifications')
            return

        coordinator_slot_info = self.get_coordinator_info()
        if not coordinator_slot_info:
            logging.debug('no coordinator info, skipping notifications')
            return

        coordinator_client = self.get_worker_client(coordinator_slot_info)
        if not coordinator_client:
            logging.debug('no coordinator client, skipping notifications')
            return

        timestamp = _epoch_time_s()
        try:
            coordinator_client.notify_hosts_updated(timestamp, update_res)
        except:
            if self._verbose >= 2:
                logging.exception('failed to notify {}[{}] of host updates'
                                  .format(coordinator_slot_info.hostname,
                                          coordinator_slot_info.local_rank))

    def _update_host_assignments(self, current_hosts):
        # Determine the slots that are already filled so we do not respawn these processes
        active_slots = set([(host, slot_info.local_rank)
                            for host, slots in self._host_assignments.items()
                            for slot_info in slots])

        # Adjust the host assignments to account for added / removed hosts
        host_assignments, host_assignments_list = self._get_host_assignments(current_hosts)

        if len(self._host_assignments) > 0:
            # Ensure that at least one previously active host is still assigned, otherwise there is no
            # way to sync the state to the new workers
            prev_hosts = self._host_assignments.keys()
            next_hosts = host_assignments.keys()
            if not prev_hosts & next_hosts:
                raise RuntimeError('No hosts from previous set remaining, unable to broadcast state.')

        self._host_assignments = host_assignments
        self._world_size = len(host_assignments_list)
        self._rendezvous.init(host_assignments_list)

        # Rank assignments map from world rank to slot info
        rank_assignments = {}
        for slot_info in host_assignments_list:
            rank_assignments[slot_info.rank] = slot_info
        self._rank_assignments = rank_assignments

        # Get the newly assigned slots that need to be started
        pending_slots = [slot_info
                         for host, slots in self._host_assignments.items()
                         for slot_info in slots
                         if (host, slot_info.local_rank) not in active_slots]
        return pending_slots

    def _get_host_assignments(self, current_hosts):
        # Adjust the host assignments to account for added / removed hosts
        host_list = [hosts.HostInfo(host, current_hosts.get_slots(host))
                     for host in current_hosts.host_assignment_order]
        host_assignments_list = hosts.get_host_assignments(host_list, self._min_num_proc, self._max_num_proc)
        host_assignments = defaultdict(list)
        for slot_info in host_assignments_list:
            host_assignments[slot_info.hostname].append(slot_info)
        return host_assignments, host_assignments_list

    def _start_worker_processes(self, pending_slots):
        for slot_info in pending_slots:
            logging.info('start worker process: {}[{}]'.format(slot_info.hostname, slot_info.local_rank))
            self._start_worker_process(slot_info)

    def _start_worker_process(self, slot_info):
        create_worker_fn = self._create_worker_fn
        shutdown_event = self._shutdown
        host_event = self._host_manager.get_host_event(slot_info.hostname)

        def run_worker():
            res = create_worker_fn(slot_info, [shutdown_event, host_event])
            exit_code, timestamp = res
            self._handle_worker_exit(slot_info, exit_code, timestamp)

        thread = threading.Thread(target=run_worker)
        thread.daemon = True
        thread.start()
        self._results.expect(thread)

    def _handle_worker_exit(self, slot_info, exit_code, timestamp):
        if not self.has_rank_assignment(slot_info.hostname, slot_info.local_rank):
            # Ignore hosts that are not assigned a rank
            logging.debug('host {} has been blacklisted, ignoring exit from local_rank={}'
                          .format(slot_info.hostname, slot_info.local_rank))
            return

        if exit_code == 0:
            rendezvous_id = self._worker_registry.record_success(slot_info.hostname, slot_info.local_rank)
        else:
            rendezvous_id = self._worker_registry.record_failure(slot_info.hostname, slot_info.local_rank)

        if self.finished() and self._worker_registry.last_rendezvous() == rendezvous_id:
            logging.debug('adding results for {}[{}]: ({}, {})'
                          .format(slot_info.hostname, slot_info.local_rank, exit_code, timestamp))
            name = '{}[{}]'.format(slot_info.hostname, slot_info.local_rank)
            self._results.add_result(name, (exit_code, timestamp))