Files @ 83dc5e1e183e
Branch filter:

Location: Morevna/src/client.py

Laman
regularly flushing the hash tree to the disk during a data transfer
import collections
import socket
import ssl
import logging as log
from datetime import datetime

import config as conf
import status
import stats
from util import Progress
from hashtree import HashTree, hash_block
from netnode import BaseConnection, NetNode, FailedConnection, LockedException, IncompatibleException
from datafile import DataFile


class DeniedConnection(Exception): pass


class Connection(BaseConnection):
	def __init__(self, host, port):
		super().__init__()
		sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)

		ssl_context = ssl.create_default_context(cafile=conf.peers)
		ssl_context.check_hostname = False
		ssl_context.load_cert_chain(conf.certfile, conf.keyfile)

		self._socket = ssl_context.wrap_socket(sock)

		try:
			self._socket.connect((host, port))
		except (ConnectionRefusedError, OSError) as e:
			log.exception(e)
			print("Couldn't connect to {0}:{1}".format(host, port))
			raise FailedConnection()
		except ssl.SSLError as e:
			log.exception(e)
			print("Error creating SSL connection to {0}:{1}".format(host, port))
			raise FailedConnection()

		self.create_networkers()
		print("Connected to {0}".format(host))


class Client(NetNode):
	def __init__(self, filename, tree_file=""):
		print(datetime.now(), "initializing...")
		super().__init__(filename, tree_file)

	def init(self, action):
		json_data = {
			"command": "init",
			"blockSize": self._tree.BLOCK_SIZE,
			"blockCount": self._tree.leaf_count,
			"version": conf.version,
			"action": action
		}
		self._outcoming.write_msg(json_data)
		json_data, bin_data = self._incoming.read_msg()
		if json_data["command"] == "deny":
			if json_data["status"] == status.incompatible.version:
				raise DeniedConnection("Incompatible client version. Consider upgrading it.")
			raise DeniedConnection()
		assert json_data["command"] == "init"
		if json_data["version"] < conf.lowest_compatible:
			raise IncompatibleException("Incompatible server version. Consider upgrading it.")

	## Asks server for node hashes to determine which are to be transferred.
	#
	# Uses a binary HashTree, where item at k is hash of items at 2k+1, 2k+2.
	#
	# Requests nodes in order of a batch DFS. Needs a stack of size O(tree_depth*batch_size). Nodes in each tree level are accessed in order.
	def negotiate(self):
		local_tree = self._tree
		blocks_to_transfer = []
		node_stack = collections.deque([0])  # root

		# determine which blocks to send
		print(datetime.now(), "negotiating:")
		progress = Progress(local_tree.leaf_count)
		while len(node_stack) > 0:
			indices = []
			for i in range(conf.batch_size.hash):
				indices.append(node_stack.pop())
				if len(node_stack) == 0: break
			self._outcoming.write_msg({"command": "req", "index": indices, "dataType": "hash"})

			json_data, bin_data = self._incoming.read_msg()
			assert json_data["index"] == indices
			assert json_data["dataType"] == "hash"
			stats.log_exchanged_node(len(indices))

			frontier = []
			for (j, i) in enumerate(indices):
				(j1, j2) = [HashTree.HASH_LEN*ji for ji in (j, j+1)]
				if local_tree.store[i] != bin_data[j1:j2]:
					# ie. 0-6 nodes, 7-14 leaves. 2*6+2<15
					if 2*i+2 < len(local_tree.store):  # inner node
						frontier.append(2*i+1)
						frontier.append(2*i+2)
					else:
						blocks_to_transfer.append(i-local_tree.leaf_start)  # leaf
						progress.p(i-local_tree.leaf_start)
			node_stack.extend(reversed(frontier))
		progress.done()

		size = stats.format_bytes(len(blocks_to_transfer)*self._tree.BLOCK_SIZE)
		print(datetime.now(), "{0} to transfer".format(size))

		return blocks_to_transfer

	def send_data(self, blocks_to_transfer):
		log.info(blocks_to_transfer)
		data_file = DataFile.open(self._filename)

		print(datetime.now(), "sending data:")
		progress = Progress(len(blocks_to_transfer))

		for k in range(0, len(blocks_to_transfer), conf.batch_size.data):
			indices = []
			blocks = []
			for j in range(conf.batch_size.data):
				if k+j >= len(blocks_to_transfer): break
				i = blocks_to_transfer[k+j]
				block = data_file.read_from(i)

				indices.append(i)
				blocks.append(block)
				log.info("block #{0}: {1}...{2}".format(i, block[:5], block[-5:]))

				progress.p(k+j)
			if indices: self._send_data(indices, blocks)
		progress.done()

		self._outcoming.write_msg({"command": "end", "action": "push"})

		log.info("closing session...")
		data_file.close()

	def pull_data(self, blocks_to_transfer, ignore_lock=False):
		if not ignore_lock:
			try:
				self._lock()
			except LockedException:
				print(
					"The file is locked. Either (a) there's another pull going on (then wait or kill it), or (b) a previous pull ended prematurely and the file is probably corrupt (then repeat pull with -f for force).")
				return
		log.info(blocks_to_transfer)
		data_file = DataFile.open(self._filename, mode="rb+")

		print(datetime.now(), "receiving data:")
		progress = Progress(len(blocks_to_transfer))

		last_flushed = datetime.now().timestamp()
		for k in range(0, len(blocks_to_transfer), conf.batch_size.data):
			indices = blocks_to_transfer[k:k+conf.batch_size.data]
			self._outcoming.write_msg({"command": "req", "index": indices, "dataType": "data"})
			json_data, bin_data = self._incoming.read_msg()
			assert json_data["command"]=="send" and json_data["index"]==indices and json_data["dataType"]=="data", json_data
			for (j, i) in enumerate(indices):
				block = bin_data[j*HashTree.BLOCK_SIZE:(j+1)*HashTree.BLOCK_SIZE]
				data_file.write_at(i, block)

				log.info("block #{0}: {1}...{2}".format(i, block[:5], block[-5:]))
				if self._tree_file:
					self._new_leaves[i+self._tree.leaf_start] = hash_block(block)

					t = datetime.now().timestamp()
					if t-last_flushed >= 60 and self._tree_file:
						self._update_tree()
						last_flushed = t

				stats.log_transferred_block()
				progress.p(k+j)
		progress.done()

		self._outcoming.write_msg({"command": "end"})

		log.info("closing session...")
		data_file.close()
		self._unlock()

		if self._tree_file:
			self._update_tree()

	def _send_data(self, indices, blocks):
		json_data = {"command": "send", "index": indices, "dataType": "data"}
		bin_data = b"".join(blocks)
		self._outcoming.write_msg(json_data, bin_data)
		stats.log_transferred_block(len(indices))
		json_data, bin_data = self._incoming.read_msg()
		assert json_data["command"]=="ack" and json_data["index"]==indices, json_data

	def set_connection(self, connection):
		(self._incoming, self._outcoming) = connection