lbry-sdk/lbrynet/blob_exchange/downloader.py
Jack Robison 6aef6a80b7
refactor downloader
split peer accumulation out, use Queues instead of locks
2019-01-31 15:39:20 -05:00

108 lines
4.5 KiB
Python

import asyncio
import typing
import logging
from lbrynet.utils import drain_tasks
from lbrynet.blob_exchange.client import request_blob
if typing.TYPE_CHECKING:
from lbrynet.conf import Config
from lbrynet.dht.node import Node
from lbrynet.dht.peer import KademliaPeer
from lbrynet.blob.blob_manager import BlobFileManager
from lbrynet.blob.blob_file import BlobFile
log = logging.getLogger(__name__)
def drain_into(a: list, b: list):
while a:
b.append(a.pop())
class BlobDownloader:
def __init__(self, loop: asyncio.BaseEventLoop, config: 'Config', blob_manager: 'BlobFileManager',
peer_queue: asyncio.Queue):
self.loop = loop
self.config = config
self.blob_manager = blob_manager
self.peer_queue = peer_queue
self.active_connections: typing.Dict['KademliaPeer', asyncio.Task] = {} # active request_blob calls
self.ignored: typing.Set['KademliaPeer'] = set()
@property
def blob_download_timeout(self):
return self.config.blob_download_timeout
@property
def peer_connect_timeout(self):
return self.config.peer_connect_timeout
@property
def max_connections(self):
return self.config.max_connections_per_download
def request_blob_from_peer(self, blob: 'BlobFile', peer: 'KademliaPeer'):
async def _request_blob():
if blob.get_is_verified():
return
success, keep_connection = await request_blob(self.loop, blob, peer.address, peer.tcp_port,
self.peer_connect_timeout, self.blob_download_timeout)
if not keep_connection and peer not in self.ignored:
self.ignored.add(peer)
log.debug("drop peer %s:%i", peer.address, peer.tcp_port)
elif keep_connection:
log.debug("keep peer %s:%i", peer.address, peer.tcp_port)
return self.loop.create_task(_request_blob())
async def new_peer_or_finished(self, blob: 'BlobFile'):
async def get_and_re_add_peers():
new_peers = await self.peer_queue.get()
self.peer_queue.put_nowait(new_peers)
tasks = [self.loop.create_task(get_and_re_add_peers()), self.loop.create_task(blob.verified.wait())]
try:
await asyncio.wait(tasks, loop=self.loop, return_when='FIRST_COMPLETED')
except asyncio.CancelledError:
drain_tasks(tasks)
async def download_blob(self, blob_hash: str, length: typing.Optional[int] = None) -> 'BlobFile':
blob = self.blob_manager.get_blob(blob_hash, length)
if blob.get_is_verified():
return blob
try:
while not blob.get_is_verified():
batch: typing.List['KademliaPeer'] = []
while not self.peer_queue.empty():
batch.extend(await self.peer_queue.get())
for peer in batch:
if peer not in self.active_connections and peer not in self.ignored:
log.info("add request %s", blob_hash[:8])
self.active_connections[peer] = self.request_blob_from_peer(blob, peer)
await self.new_peer_or_finished(blob)
log.info("new peer or finished %s", blob_hash[:8])
to_re_add = list(set(filter(lambda peer: peer not in self.ignored, batch)))
if to_re_add:
self.peer_queue.put_nowait(to_re_add)
log.info("finished %s", blob_hash[:8])
while self.active_connections:
peer, task = self.active_connections.popitem()
if task and not task.done():
task.cancel()
return blob
except asyncio.CancelledError:
while self.active_connections:
peer, task = self.active_connections.popitem()
if task and not task.done():
task.cancel()
raise
async def download_blob(loop, config: 'Config', blob_manager: 'BlobFileManager', node: 'Node',
blob_hash: str) -> 'BlobFile':
search_queue = asyncio.Queue(loop=loop)
search_queue.put_nowait(blob_hash)
peer_queue, accumulate_task = node.accumulate_peers(search_queue)
downloader = BlobDownloader(loop, config, blob_manager, peer_queue)
try:
return await downloader.download_blob(blob_hash)
finally:
if accumulate_task and not accumulate_task.done():
accumulate_task.cancel()