instance1 / Instance.py
ChandimaPrabath's picture
init
dd2f10e
raw
history blame
13.3 kB
import os
import requests
import json
import time
from threading import Thread, Event
from requests.exceptions import RequestException
from tqdm import tqdm
import logging
CACHE_DIR = os.getenv("CACHE_DIR")
download_progress = {}
class Instance:
def __init__(self, id, url, cache_dir, token, repo, load_balancer_api, max_retries=20, initial_delay=1):
self.version = "0.0.0.1 Alpha"
self.id = id
self.url = url
self.CACHE_DIR = cache_dir
self.TOKEN = token
self.REPO = repo
self.FILM_STORE = {}
self.TV_STORE = {}
self.download_threads = {}
self.file_structure = None
self.load_balancer_api = load_balancer_api
self.max_retries = max_retries
self.initial_delay = initial_delay
self.last_report_time = time.time()
self.re_register_event = Event()
# Ensure CACHE_DIR exists
if not os.path.exists(self.CACHE_DIR):
os.makedirs(self.CACHE_DIR)
self.register_to_load_balancer()
self.reload_file_structure()
registration_thread = Thread(target=self.monitor_registration)
registration_thread.daemon = True
registration_thread.start()
# Start the thread to re-index every 2 minutes
indexer_thread = Thread(target=self.get_file_structure_periodically)
indexer_thread.daemon = True
indexer_thread.start()
def reload_file_structure(self):
"""Runs the indexer and loads the file structure from INDEX_FILE."""
self.file_structure = self.load_balancer_api.get_file_structure()
logging.info("File structure reloaded successfully.")
def get_file_structure_periodically(self):
"""Periodically reruns the indexer and reloads the file structure."""
while True:
time.sleep(300) # Wait for 5 minutes
logging.info("Re-running indexer and reloading file structure.")
self.reload_file_structure()
def compile_report(self):
self.last_report_time = time.time() # Update the last report time
cache_size = self.get_cache_size()
report = {
"instance_id": self.id,
"instance_url": self.url,
"film_store": self.FILM_STORE,
"tv_store": self.TV_STORE,
"cache_size": cache_size
}
return report
def register_to_load_balancer(self):
result = self.load_balancer_api.register_instance(self.id, self.url)
if result is not None:
logging.info(f'Registered instance {self.id} to load balancer.')
else:
logging.error(f'Failed to register instance {self.id} to load balancer.')
def monitor_registration(self):
while True:
if time.time() - self.last_report_time > 60: # Check if 1 minute has passed
logging.info('1 minute passed since last report. Re-registering...')
self.register_to_load_balancer()
self.last_report_time = time.time() # Reset the last report time
time.sleep(30) # Check every 30 seconds
def get_cache_size(self):
total_size = 0
for dirpath, dirnames, filenames in os.walk(CACHE_DIR):
for f in filenames:
fp = os.path.join(dirpath, f)
total_size += os.path.getsize(fp)
return {"cache_size": f"{total_size / (1024 * 1024 * 1024):.2f} GB"}
@staticmethod
def read_json(file_path):
if os.path.exists(file_path):
with open(file_path, 'r') as json_file:
return json.load(json_file)
return {}
def download_film(self, file_url, token, cache_path, film_id, title, chunk_size=100 * 1024 * 1024):
"""
Downloads a file from the specified URL and saves it to the cache path.
Tracks the download progress.
Args:
file_url (str): The URL of the file to download.
token (str): The authorization token for the request.
cache_path (str): The path to save the downloaded file.
film_id (str): Unique identifier for the film download.
title (str): The title of the film.
chunk_size (int): Size of each chunk to download.
"""
print(f"Downloading file from URL: {file_url} to {cache_path}")
headers = {'Authorization': f'Bearer {token}'}
try:
response = requests.get(file_url, headers=headers, stream=True)
response.raise_for_status()
total_size = int(response.headers.get('content-length', 0))
download_progress[film_id] = {"total": total_size, "downloaded": 0, "status": "Downloading", "start_time": time.time()}
os.makedirs(os.path.dirname(cache_path), exist_ok=True)
with open(cache_path, 'wb') as file, tqdm(total=total_size, unit='B', unit_scale=True, desc=cache_path) as pbar:
for data in response.iter_content(chunk_size=chunk_size):
file.write(data)
pbar.update(len(data))
download_progress[film_id]["downloaded"] += len(data)
print(f'File cached to {cache_path} successfully.')
self.FILM_STORE[title] = cache_path
download_progress[film_id]["status"] = "Completed"
except RequestException as e:
print(f"Error downloading file: {e}")
download_progress[film_id]["status"] = "Failed"
except IOError as e:
print(f"Error writing file {cache_path}: {e}")
download_progress[film_id]["status"] = "Failed"
finally:
if download_progress[film_id]["status"] != "Downloading":
download_progress[film_id]["end_time"] = time.time()
@staticmethod
def get_download_progress(id):
"""
Gets the download progress for a specific film.
Args:
film_id (str): The unique identifier for the film download.
Returns:
dict: A dictionary containing the total size, downloaded size, progress percentage, status, and ETA.
"""
if id in download_progress:
total = download_progress[id]["total"]
downloaded = download_progress[id]["downloaded"]
status = download_progress[id].get("status", "In Progress")
progress = (downloaded / total) * 100 if total > 0 else 0
eta = None
if status == "Downloading" and downloaded > 0:
elapsed_time = time.time() - download_progress[id]["start_time"]
estimated_total_time = elapsed_time * (total / downloaded)
eta = estimated_total_time - elapsed_time
elif status == "Completed":
eta = 0
return {"total": total, "downloaded": downloaded, "progress": progress, "status": status, "eta": eta}
return {"total": 0, "downloaded": 0, "progress": 0, "status": "Not Found", "eta": None}
def download_episode(self, file_url, token, cache_path, episode_id, title, chunk_size=100 * 1024 * 1024):
"""
Downloads a file from the specified URL and saves it to the cache path.
Tracks the download progress.
Args:
file_url (str): The URL of the file to download.
token (str): The authorization token for the request.
cache_path (str): The path to save the downloaded file.
episode_id (str): Unique identifier for the film download.
title (str): The title of the film.
chunk_size (int): Size of each chunk to download.
"""
print(f"Downloading file from URL: {file_url} to {cache_path}")
headers = {'Authorization': f'Bearer {token}'}
try:
response = requests.get(file_url, headers=headers, stream=True)
response.raise_for_status()
total_size = int(response.headers.get('content-length', 0))
download_progress[episode_id] = {"total": total_size, "downloaded": 0, "status": "Downloading", "start_time": time.time()}
os.makedirs(os.path.dirname(cache_path), exist_ok=True)
with open(cache_path, 'wb') as file, tqdm(total=total_size, unit='B', unit_scale=True, desc=cache_path) as pbar:
for data in response.iter_content(chunk_size=chunk_size):
file.write(data)
pbar.update(len(data))
download_progress[episode_id]["downloaded"] += len(data)
print(f'File cached to {cache_path} successfully.')
self.update_tv_store(title, cache_path)
download_progress[episode_id]["status"] = "Completed"
except RequestException as e:
print(f"Error downloading file: {e}")
download_progress[episode_id]["status"] = "Failed"
except IOError as e:
print(f"Error writing file {cache_path}: {e}")
download_progress[episode_id]["status"] = "Failed"
finally:
if download_progress[episode_id]["status"] != "Downloading":
download_progress[episode_id]["end_time"] = time.time()
def update_tv_store(self, title, cache_path):
"""
Updates the TV store JSON with the new file, organizing by title, season, and episode.
Args:
title (str): The title of the TV show.
cache_path (str): The local path where the file is saved.
"""
# Extract season and episode information from the cache_path
season_part = os.path.basename(os.path.dirname(cache_path)) # Extracts 'Season 1'
episode_part = os.path.basename(cache_path) # Extracts 'Grand Blue Dreaming - S01E01 - Deep Blue HDTV-720p.mp4'
# Create the structure if not already present
if title not in self.TV_STORE:
self.TV_STORE[title] = {}
if season_part not in self.TV_STORE[title]:
self.TV_STORE[title][season_part] = {}
# Assuming episode_part is unique for each episode within a season
self.TV_STORE[title][season_part][episode_part] = cache_path
print(f'TV store updated with {title}, {season_part}, {episode_part}.')
def load_json(self, file_path):
"""Load JSON data from a file."""
with open(file_path, 'r') as file:
return json.load(file)
def find_movie_path(self, title):
"""Find the path of the movie in the JSON data based on the title."""
for directory in self.file_structure:
if directory['type'] == 'directory' and directory['path'] == 'films':
for sub_directory in directory['contents']:
if sub_directory['type'] == 'directory':
for item in sub_directory['contents']:
if item['type'] == 'file' and title.lower() in item['path'].lower():
return item['path']
return None
def find_tv_path(self, title):
"""Find the path of the TV show in the JSON data based on the title."""
for directory in self.file_structure:
if directory['type'] == 'directory' and directory['path'] == 'tv':
for sub_directory in directory['contents']:
if sub_directory['type'] == 'directory' and title.lower() in sub_directory['path'].lower():
return sub_directory['path']
return None
def get_tv_structure(self, title):
"""Find the path of the TV show in the JSON data based on the title."""
for directory in self.file_structure:
if directory['type'] == 'directory' and directory['path'] == 'tv':
for sub_directory in directory['contents']:
if sub_directory['type'] == 'directory' and title.lower() in sub_directory['path'].lower():
return sub_directory
return None
def get_film_id(self, title):
"""Generate a film ID based on the title."""
return title.replace(" ", "_").lower()
def bytes_to_human_readable(self, num, suffix="B"):
for unit in ["", "K", "M", "G", "T", "P", "E", "Z"]:
if abs(num) < 1024.0:
return f"{num:3.1f} {unit}{suffix}"
num /= 1024.0
return f"{num:.1f} Y{suffix}"
def encode_episodeid(self, title, season, episode):
return f"{title}_{season}_{episode}"
def register_to_load_balancer(self):
retries = 0
delay = self.initial_delay
max_delay = 120
while True:
try:
result = self.load_balancer_api.register_instance(self.id, self.url)
if result:
logging.info(f'Successfully registered instance {self.id} to load balancer.')
return result
except Exception as e:
logging.error(f'Error during registration: {e}')
retries += 1
logging.warning(f'Attempt {retries} to register instance {self.id} failed. Retrying in {delay} seconds...')
time.sleep(delay)
delay = min(delay * 2, max_delay) # Exponential backoff with maximum delay