from __future__ import annotations import asyncio import logging import uuid from collections import defaultdict from collections.abc import Hashable from dask.utils import parse_timedelta from distributed.client import Client from distributed.utils import TimeoutError, log_errors, wait_for from distributed.worker import get_worker logger = logging.getLogger(__name__) class MultiLockExtension: """An extension for the scheduler to manage MultiLocks This adds the following routes to the scheduler * multi_lock_acquire * multi_lock_release The approach is to maintain `self.locks` that maps a lock (unique name given to `MultiLock(names=, ...)` at creation) to a list of users (instances of `MultiLock`) that "requests" the lock. Additionally, `self.requests` maps a user to its requested locks and `self.requests_left` maps a user to the number of locks still need. Every time a user `x` gets to the front in `self.locks[name] = [x, ...]` it means that `x` now holds the lock `name` and when it holds all the requested locks `acquire()` can return. Finally, `self.events` contains all the events users are waiting on to finish. """ def __init__(self, scheduler): self.scheduler = scheduler self.locks = defaultdict(list) # lock -> users self.requests = {} # user -> locks self.requests_left = {} # user -> locks still needed self.events = {} self.scheduler.handlers.update( {"multi_lock_acquire": self.acquire, "multi_lock_release": self.release} ) def _request_locks(self, locks: list[str], id: Hashable, num_locks: int) -> bool: """Request locks Parameters ---------- locks: List[str] Names of the locks to request. id: Hashable Identifier of the `MultiLock` instance requesting the locks. num_locks: int Number of locks in `locks` requesting Return ------ result: bool Whether `num_locks` requested locks are free immediately or not. """ assert id not in self.requests self.requests[id] = set(locks) assert len(locks) >= num_locks and num_locks > 0 self.requests_left[id] = num_locks locks = sorted(locks, key=lambda x: len(self.locks[x])) for i, lock in enumerate(locks): self.locks[lock].append(id) if len(self.locks[lock]) == 1: # The lock was free self.requests_left[id] -= 1 if self.requests_left[id] == 0: # Got all locks needed # Since we got all locks need, we can remove the rest of the requests self.requests[id] -= set(locks[i + 1 :]) return True return False def _refain_locks(self, locks, id): """Cancel/release previously requested/acquired locks Parameters ---------- locks: List[str] Names of the locks to refain. id: Hashable Identifier of the `MultiLock` instance refraining the locks. """ waiters_ready = set() for lock in locks: if self.locks[lock][0] == id: self.locks[lock].pop(0) if self.locks[lock]: new_first = self.locks[lock][0] self.requests_left[new_first] -= 1 if self.requests_left[new_first] <= 0: # Notice, `self.requests_left[new_first]` might go below zero # if more locks are freed than requested. self.requests_left[new_first] = 0 waiters_ready.add(new_first) else: self.locks[lock].remove(id) assert id not in self.locks[lock] del self.requests[id] del self.requests_left[id] for waiter in waiters_ready: self.scheduler.loop.add_callback(self.events[waiter].set) @log_errors async def acquire(self, locks=None, id=None, timeout=None, num_locks=None): if not self._request_locks(locks, id, num_locks): assert id not in self.events event = asyncio.Event() self.events[id] = event future = event.wait() if timeout is not None: future = wait_for(future, timeout) try: await future except TimeoutError: self._refain_locks(locks, id) return False finally: del self.events[id] # At this point `id` acquired all `locks` assert self.requests_left[id] == 0 return True @log_errors def release(self, id=None): self._refain_locks(self.requests[id], id) class MultiLock: """Distributed Centralized Lock Parameters ---------- names Names of the locks to acquire. Choosing the same name allows two disconnected processes to coordinate a lock. client Client to use for communication with the scheduler. If not given, the default global client will be used. Examples -------- >>> lock = MultiLock(['x', 'y']) # doctest: +SKIP >>> lock.acquire(timeout=1) # doctest: +SKIP >>> # do things with protected resource 'x' and 'y' >>> lock.release() # doctest: +SKIP """ def __init__(self, names: list[str] | None = None, client: Client | None = None): try: self.client = client or Client.current() except ValueError: # Initialise new client self.client = get_worker().client self.names = names or [] self.id = uuid.uuid4().hex self._locked = False def acquire(self, blocking=True, timeout=None, num_locks=None): """Acquire the lock Parameters ---------- blocking : bool, optional If false, don't wait on the lock in the scheduler at all. timeout : string or number or timedelta, optional Seconds to wait on the lock in the scheduler. This does not include local coroutine time, network transfer time, etc.. It is forbidden to specify a timeout when blocking is false. Instead of number of seconds, it is also possible to specify a timedelta in string format, e.g. "200ms". num_locks : int, optional Number of locks needed. If None, all locks are needed Examples -------- >>> lock = MultiLock(['x', 'y']) # doctest: +SKIP >>> lock.acquire(timeout="1s") # doctest: +SKIP Returns ------- True or False whether or not it successfully acquired the lock """ timeout = parse_timedelta(timeout) if not blocking: if timeout is not None: raise ValueError("can't specify a timeout for a non-blocking call") timeout = 0 result = self.client.sync( self.client.scheduler.multi_lock_acquire, locks=self.names, id=self.id, timeout=timeout, num_locks=num_locks or len(self.names), ) self._locked = True return result def release(self): """Release the lock if already acquired""" if not self.locked(): raise ValueError("Lock is not yet acquired") ret = self.client.sync(self.client.scheduler.multi_lock_release, id=self.id) self._locked = False return ret def locked(self): return self._locked def __enter__(self): self.acquire() return self def __exit__(self, exc_type, exc_value, traceback): self.release() async def __aenter__(self): await self.acquire() return self async def __aexit__(self, exc_type, exc_value, traceback): await self.release() def __reduce__(self): return (type(self), (self.names,))