chnage service discovery
This commit is contained in:
+183
-227
@@ -1,40 +1,41 @@
|
||||
"""LeaderClient — locates meta-core and exposes its URLs.
|
||||
|
||||
Since meta-discovery v1 this is a thin adapter over ``MetaCoreLocator``
|
||||
(``meshdisco.py``): meta-core is found by UDP announce instead of by reading
|
||||
``/meta-core/locks/kv-leader.info``, so meta-stremio no longer needs the
|
||||
``/meta-core`` volume mounted at all.
|
||||
|
||||
The public surface is deliberately unchanged so ``LeaderStorage`` did not have
|
||||
to move: ``get_leader_info``, ``get_urls``, ``get_api_url``,
|
||||
``get_webdav_url*``, ``wait_for_leader``, ``start_watching`` / ``on_change``
|
||||
(the watchdog on the lock directory became "an announce arrived carrying a
|
||||
different apiUrl", feeding the same callback).
|
||||
|
||||
Precedence is the pin: when ``meta_core_url`` (``META_CORE_URL``) is set it
|
||||
always wins and the wire is never consulted for core selection. See
|
||||
docs/project-architecture/service-discovery.md.
|
||||
|
||||
The name "leader" is legacy: meta-core's flock election is vestigial, and
|
||||
``redis_url`` has been empty since the api-mediated-access lockdown — every
|
||||
read goes over HTTP.
|
||||
"""
|
||||
LeaderClient - Client for reading leader info from meta-core
|
||||
|
||||
This replaces the old LeaderDiscovery.py by delegating leader election
|
||||
to meta-core (Go sidecar) and just reading the results.
|
||||
|
||||
Features:
|
||||
- Reads kv-leader.info file for leader URLs
|
||||
- Calls meta-core /urls API for current leader URLs
|
||||
- Watches for leader changes via file system watcher
|
||||
"""
|
||||
from __future__ import annotations
|
||||
|
||||
import os
|
||||
import json
|
||||
import threading
|
||||
import os
|
||||
import time
|
||||
import urllib.request
|
||||
from dataclasses import dataclass
|
||||
from pathlib import Path
|
||||
from typing import Optional, Callable
|
||||
from urllib.request import urlopen, Request
|
||||
from urllib.error import URLError
|
||||
from typing import Callable, List, Optional
|
||||
|
||||
# Try to import watchdog for file watching (optional)
|
||||
try:
|
||||
from watchdog.observers import Observer
|
||||
from watchdog.events import FileSystemEventHandler
|
||||
WATCHDOG_AVAILABLE = True
|
||||
except ImportError:
|
||||
WATCHDOG_AVAILABLE = False
|
||||
Observer = None
|
||||
FileSystemEventHandler = object
|
||||
from .meshdisco import MetaCoreLocator
|
||||
|
||||
|
||||
@dataclass
|
||||
class LeaderLockInfo:
|
||||
"""Leader lock file content format (kv-leader.info).
|
||||
Written by meta-core leader, read by other services.
|
||||
"""meta-core's identity and URLs.
|
||||
|
||||
Historically the parsed kv-leader.info + /urls pair; now assembled from the
|
||||
announce (or from /urls when meta-core is pinned).
|
||||
"""
|
||||
hostname: str
|
||||
base_url: str
|
||||
@@ -53,140 +54,133 @@ class URLsResponse:
|
||||
base_url: str
|
||||
api_url: str
|
||||
redis_url: str
|
||||
webdav_url: str # External WebDAV URL (via nginx/HTTPS)
|
||||
webdav_url_internal: str # Internal WebDAV URL (direct to port 9000)
|
||||
webdav_url: str
|
||||
webdav_url_internal: str
|
||||
is_leader: bool
|
||||
|
||||
|
||||
class LockFileHandler(FileSystemEventHandler):
|
||||
"""File system event handler for lock file changes."""
|
||||
|
||||
def __init__(self, callback: Callable[[], None], lock_filename: str):
|
||||
self.callback = callback
|
||||
self.lock_filename = lock_filename
|
||||
|
||||
def on_modified(self, event):
|
||||
if event.src_path.endswith(self.lock_filename):
|
||||
self.callback()
|
||||
|
||||
def on_created(self, event):
|
||||
if event.src_path.endswith(self.lock_filename):
|
||||
self.callback()
|
||||
|
||||
def on_deleted(self, event):
|
||||
if event.src_path.endswith(self.lock_filename):
|
||||
self.callback()
|
||||
|
||||
|
||||
class LeaderClient:
|
||||
"""
|
||||
LeaderClient for meta-stremio.
|
||||
|
||||
Reads leader info from kv-leader.info file written by meta-core.
|
||||
"""
|
||||
"""Locates meta-core over UDP and resolves its URL set."""
|
||||
|
||||
def __init__(
|
||||
self,
|
||||
meta_core_path: str,
|
||||
meta_core_url: Optional[str] = None
|
||||
):
|
||||
self.meta_core_path = meta_core_path
|
||||
self.meta_core_url = meta_core_url
|
||||
meta_core_path: Optional[str] = None,
|
||||
meta_core_url: Optional[str] = None,
|
||||
service_name: str = "meta-stremio",
|
||||
base_url: Optional[str] = None,
|
||||
version: str = "",
|
||||
) -> None:
|
||||
# meta_core_path is accepted and ignored — kept so existing callers
|
||||
# construct unchanged. Nothing reads the volume any more.
|
||||
self._meta_core_path = meta_core_path
|
||||
self._meta_core_url = (meta_core_url or "").strip() or None
|
||||
self._leader_info: Optional[LeaderLockInfo] = None
|
||||
self._on_change_callbacks: List[Callable[[], None]] = []
|
||||
self._started = False
|
||||
|
||||
self.info_file_path = os.path.join(meta_core_path, 'locks', 'kv-leader.info')
|
||||
self.leader_info: Optional[LeaderLockInfo] = None
|
||||
|
||||
self._file_observer = None
|
||||
self._on_change_callbacks: list[Callable[[], None]] = []
|
||||
self._stop_event = threading.Event()
|
||||
|
||||
# URL caching
|
||||
# Only used on the pinned path, where the URLs still come from an HTTP
|
||||
# GET /urls. A discovered core carries them in the announce itself.
|
||||
self._cached_urls: Optional[URLsResponse] = None
|
||||
self._urls_cache_time: float = 0
|
||||
self._urls_cache_ttl: float = 5.0 # 5 seconds
|
||||
self._urls_cache_time: float = 0.0
|
||||
self._urls_cache_ttl: float = 5.0
|
||||
|
||||
def _get_api_url_from_file(self) -> Optional[str]:
|
||||
"""Read API URL from file (plain text format)."""
|
||||
try:
|
||||
with open(self.info_file_path, 'r') as f:
|
||||
content = f.read()
|
||||
return content.strip() or None
|
||||
except FileNotFoundError:
|
||||
print(f"[LeaderClient] Leader info file not found: {self.info_file_path}")
|
||||
return None
|
||||
except Exception as e:
|
||||
print(f"[LeaderClient] Error reading API URL from file: {e}")
|
||||
return None
|
||||
self._locator = MetaCoreLocator(
|
||||
service_name=service_name,
|
||||
base_url=base_url,
|
||||
version=version,
|
||||
meta_core_url=self._meta_core_url,
|
||||
)
|
||||
self._locator.on_change(self._on_locator_change)
|
||||
|
||||
# -- lifecycle ---------------------------------------------------------
|
||||
|
||||
def _ensure_started(self) -> None:
|
||||
if self._started:
|
||||
return
|
||||
self._started = True
|
||||
self._locator.start()
|
||||
|
||||
def _on_locator_change(self) -> None:
|
||||
self._cached_urls = None
|
||||
self._urls_cache_time = 0.0
|
||||
self._notify_change()
|
||||
|
||||
# -- url resolution ----------------------------------------------------
|
||||
|
||||
def _assert_no_redis_url(self, redis_url: Optional[str]) -> None:
|
||||
"""Guard against a rollback reintroducing direct Redis exposure.
|
||||
|
||||
api-mediated-access PR D removed redis_url; ALLOW_LEGACY_REDIS_URL=1
|
||||
downgrades this to a warning during a deliberate temporary rollback.
|
||||
"""
|
||||
if not redis_url:
|
||||
return
|
||||
msg = (
|
||||
"meta-core still publishes redisUrl; direct Redis access was retired "
|
||||
"by the api-mediated-access lockdown. Verify meta-core version."
|
||||
)
|
||||
if os.environ.get("ALLOW_LEGACY_REDIS_URL") == "1":
|
||||
print(f"[LeaderClient] WARNING: {msg}")
|
||||
else:
|
||||
raise RuntimeError(msg)
|
||||
|
||||
def _fetch_urls(self, api_url: str) -> Optional[URLsResponse]:
|
||||
"""Fetch URLs from meta-core /urls API with caching."""
|
||||
import time
|
||||
|
||||
# Check cache
|
||||
now = time.time()
|
||||
if self._cached_urls and (now - self._urls_cache_time) < self._urls_cache_ttl:
|
||||
return self._cached_urls
|
||||
|
||||
try:
|
||||
url = f"{api_url}/urls"
|
||||
req = Request(url, headers={'Accept': 'application/json'})
|
||||
with urlopen(req, timeout=5) as response:
|
||||
if response.status != 200:
|
||||
print(f"[LeaderClient] Failed to fetch URLs: {response.status}")
|
||||
return None
|
||||
|
||||
data = json.loads(response.read().decode())
|
||||
|
||||
# api-mediated-access PR D: redisUrl is no longer published by
|
||||
# meta-core. If it shows up, an old build snuck back in —
|
||||
# fail loudly so the rollback can't go unnoticed. Set
|
||||
# ALLOW_LEGACY_REDIS_URL=1 to downgrade to a warning during
|
||||
# a deliberate temporary rollback.
|
||||
if data.get('redisUrl'):
|
||||
msg = (
|
||||
"[LeaderClient] meta-core still publishes redisUrl; "
|
||||
"direct Redis access was retired by the api-mediated-access "
|
||||
"lockdown. Verify meta-core version."
|
||||
)
|
||||
if os.environ.get("ALLOW_LEGACY_REDIS_URL") == "1":
|
||||
print("WARNING: " + msg)
|
||||
else:
|
||||
raise RuntimeError(msg)
|
||||
self._cached_urls = URLsResponse(
|
||||
hostname=data.get('hostname', ''),
|
||||
base_url=data.get('baseUrl', ''),
|
||||
api_url=data.get('apiUrl', ''),
|
||||
redis_url=data.get('redisUrl', ''),
|
||||
webdav_url=data.get('webdavUrl', ''),
|
||||
webdav_url_internal=data.get('webdavUrlInternal', ''),
|
||||
is_leader=data.get('isLeader', False)
|
||||
)
|
||||
self._urls_cache_time = now
|
||||
return self._cached_urls
|
||||
|
||||
except URLError as e:
|
||||
req = urllib.request.Request(
|
||||
f"{api_url.rstrip('/')}/urls", headers={"Accept": "application/json"}
|
||||
)
|
||||
with urllib.request.urlopen(req, timeout=5) as resp:
|
||||
data = json.loads(resp.read().decode("utf-8"))
|
||||
self._assert_no_redis_url(data.get("redisUrl"))
|
||||
self._cached_urls = URLsResponse(
|
||||
hostname=data.get("hostname", ""),
|
||||
base_url=data.get("baseUrl", ""),
|
||||
api_url=data.get("apiUrl", ""),
|
||||
redis_url=data.get("redisUrl", "") or "",
|
||||
webdav_url=data.get("webdavUrl", ""),
|
||||
webdav_url_internal=data.get("webdavUrlInternal", ""),
|
||||
is_leader=bool(data.get("isLeader", True)),
|
||||
)
|
||||
self._urls_cache_time = now
|
||||
return self._cached_urls
|
||||
except Exception as e:
|
||||
print(f"[LeaderClient] Error calling /urls API: {e}")
|
||||
return None
|
||||
except Exception as e:
|
||||
print(f"[LeaderClient] Error fetching URLs: {e}")
|
||||
|
||||
def get_urls(self) -> Optional[URLsResponse]:
|
||||
"""Resolve meta-core's URLs.
|
||||
|
||||
Straight from the announce when discovered (no HTTP hop needed); over
|
||||
HTTP when meta-core is pinned.
|
||||
"""
|
||||
self._ensure_started()
|
||||
|
||||
wire = self._locator.get_urls()
|
||||
if wire and not self._meta_core_url:
|
||||
return URLsResponse(
|
||||
hostname=wire.get("hostname", ""),
|
||||
base_url=wire.get("baseUrl", ""),
|
||||
api_url=wire.get("apiUrl", ""),
|
||||
redis_url="",
|
||||
webdav_url=wire.get("webdavUrl", ""),
|
||||
webdav_url_internal=wire.get("webdavUrlInternal", ""),
|
||||
is_leader=True,
|
||||
)
|
||||
|
||||
api_url = self._locator.get_api_url()
|
||||
if not api_url:
|
||||
return None
|
||||
return self._fetch_urls(api_url)
|
||||
|
||||
def get_leader_info(self) -> Optional[LeaderLockInfo]:
|
||||
"""Read leader info from file and /urls API."""
|
||||
try:
|
||||
# Read API URL from file (plain text)
|
||||
api_url = self._get_api_url_from_file()
|
||||
if not api_url:
|
||||
return None
|
||||
|
||||
# Fetch full info from /urls API
|
||||
urls = self._fetch_urls(api_url)
|
||||
urls = self.get_urls()
|
||||
if not urls:
|
||||
return None
|
||||
|
||||
# Convert URLsResponse to LeaderLockInfo
|
||||
import time
|
||||
self.leader_info = LeaderLockInfo(
|
||||
self._leader_info = LeaderLockInfo(
|
||||
hostname=urls.hostname,
|
||||
base_url=urls.base_url,
|
||||
api_url=urls.api_url,
|
||||
@@ -194,145 +188,107 @@ class LeaderClient:
|
||||
webdav_url=urls.webdav_url,
|
||||
webdav_url_internal=urls.webdav_url_internal,
|
||||
timestamp=int(time.time() * 1000),
|
||||
pid=0 # Unknown for remote leader
|
||||
pid=0,
|
||||
)
|
||||
|
||||
return self.leader_info
|
||||
|
||||
return self._leader_info
|
||||
except Exception as e:
|
||||
print(f"[LeaderClient] Error reading leader info: {e}")
|
||||
print(f"[LeaderClient] Failed to read leader info: {e}")
|
||||
return None
|
||||
|
||||
def get_redis_url(self) -> Optional[str]:
|
||||
"""Get Redis URL from leader info."""
|
||||
info = self.get_leader_info()
|
||||
return info.redis_url if info else None
|
||||
|
||||
def get_webdav_url(self) -> Optional[str]:
|
||||
"""Get WebDAV URL from leader info (external, via nginx/HTTPS)."""
|
||||
info = self.get_leader_info()
|
||||
return info.webdav_url if info else None
|
||||
|
||||
def get_webdav_url_internal(self) -> Optional[str]:
|
||||
"""Get internal WebDAV URL from leader info (direct to port 9000).
|
||||
Use this for container-to-container communication.
|
||||
"""
|
||||
"""Internal WebDAV URL — use for container-to-container access."""
|
||||
info = self.get_leader_info()
|
||||
return info.webdav_url_internal if info else None
|
||||
|
||||
def get_api_url(self) -> Optional[str]:
|
||||
"""Get meta-core API URL from leader info."""
|
||||
info = self.get_leader_info()
|
||||
return info.api_url if info else None
|
||||
|
||||
def get_urls(self) -> Optional[URLsResponse]:
|
||||
"""Call meta-core /urls API to get current URLs."""
|
||||
# First try using configured meta_core_url
|
||||
api_url = self.meta_core_url
|
||||
|
||||
# Fall back to reading from file
|
||||
if not api_url:
|
||||
api_url = self._get_api_url_from_file()
|
||||
|
||||
if not api_url:
|
||||
print("[LeaderClient] No meta-core API URL available")
|
||||
return None
|
||||
|
||||
return self._fetch_urls(api_url)
|
||||
|
||||
def wait_for_leader(self, timeout_ms: int = 30000) -> LeaderLockInfo:
|
||||
"""Wait for leader info to be available."""
|
||||
import time
|
||||
start_time = time.time()
|
||||
poll_interval = 0.5 # 500ms
|
||||
"""Block until meta-core is reachable.
|
||||
|
||||
while (time.time() - start_time) * 1000 < timeout_ms:
|
||||
Probes immediately then every 500ms, matching the previous
|
||||
file-polling loop's timing.
|
||||
"""
|
||||
self._ensure_started()
|
||||
timeout_s = None if timeout_ms <= 0 else timeout_ms / 1000.0
|
||||
deadline = None if timeout_s is None else time.time() + timeout_s
|
||||
|
||||
if self._locator.wait_for_core(timeout_s) is None:
|
||||
raise TimeoutError(f"No meta-core found within {timeout_ms}ms")
|
||||
|
||||
while deadline is None or time.time() < deadline:
|
||||
info = self.get_leader_info()
|
||||
if info:
|
||||
print(f"[LeaderClient] Leader found: {info.hostname} at {info.redis_url}")
|
||||
print(f"[LeaderClient] meta-core found: {info.hostname} at {info.api_url}")
|
||||
return info
|
||||
time.sleep(0.5)
|
||||
raise TimeoutError(f"No meta-core found within {timeout_ms}ms")
|
||||
|
||||
print("[LeaderClient] Waiting for leader...")
|
||||
time.sleep(poll_interval)
|
||||
|
||||
raise TimeoutError(f"No leader found within {timeout_ms}ms")
|
||||
# -- change notification ----------------------------------------------
|
||||
|
||||
def start_watching(self) -> None:
|
||||
"""Start watching for leader changes."""
|
||||
if not WATCHDOG_AVAILABLE:
|
||||
print("[LeaderClient] watchdog not available, file watching disabled")
|
||||
return
|
||||
"""Previously a watchdog on the lock directory.
|
||||
|
||||
try:
|
||||
lock_dir = os.path.dirname(self.info_file_path)
|
||||
|
||||
handler = LockFileHandler(
|
||||
callback=self._on_file_change,
|
||||
lock_filename='kv-leader.info'
|
||||
)
|
||||
|
||||
self._file_observer = Observer()
|
||||
self._file_observer.schedule(handler, lock_dir, recursive=False)
|
||||
self._file_observer.start()
|
||||
print(f"[LeaderClient] Watching for leader changes in {lock_dir}")
|
||||
|
||||
except Exception as e:
|
||||
print(f"[LeaderClient] Could not start file watcher: {e}")
|
||||
Discovery is always listening, so this only has to make sure the node
|
||||
is running.
|
||||
"""
|
||||
self._ensure_started()
|
||||
|
||||
def stop_watching(self) -> None:
|
||||
"""Stop watching for leader changes."""
|
||||
if self._file_observer:
|
||||
self._file_observer.stop()
|
||||
self._file_observer.join(timeout=2.0)
|
||||
self._file_observer = None
|
||||
print("[LeaderClient] Stopped watching for leader changes")
|
||||
"""No-op: the locator keeps listening until close()."""
|
||||
|
||||
def _on_file_change(self) -> None:
|
||||
"""Handle file change event."""
|
||||
print("[LeaderClient] Leader info changed, invalidating cache...")
|
||||
# Invalidate cache to force fresh API call
|
||||
self._cached_urls = None
|
||||
self._urls_cache_time = 0
|
||||
|
||||
self.get_leader_info()
|
||||
self._notify_change()
|
||||
|
||||
def on_change(self, callback: Callable[[], None]) -> 'LeaderClient':
|
||||
"""Register callback for leader changes."""
|
||||
def on_change(self, callback: Callable[[], None]) -> "LeaderClient":
|
||||
self._on_change_callbacks.append(callback)
|
||||
return self
|
||||
|
||||
def _notify_change(self) -> None:
|
||||
"""Notify all change callbacks."""
|
||||
for callback in self._on_change_callbacks:
|
||||
print("[LeaderClient] meta-core changed, invalidating cache...")
|
||||
for cb in list(self._on_change_callbacks):
|
||||
try:
|
||||
callback()
|
||||
cb()
|
||||
except Exception as e:
|
||||
print(f"[LeaderClient] Error in change callback: {e}")
|
||||
|
||||
def get_cached_leader_info(self) -> Optional[LeaderLockInfo]:
|
||||
"""Get cached leader info (without re-reading file)."""
|
||||
return self.leader_info
|
||||
return self._leader_info
|
||||
|
||||
# -- nav menu ----------------------------------------------------------
|
||||
|
||||
def neighbors(self) -> List[dict]:
|
||||
"""Neighbours for the nav menu (one row per service name)."""
|
||||
self._ensure_started()
|
||||
return self._locator.neighbors()
|
||||
|
||||
def self_announce(self) -> dict:
|
||||
"""This service's own announce, so the menu can show itself."""
|
||||
self._ensure_started()
|
||||
return self._locator.self_announce()
|
||||
|
||||
def close(self) -> None:
|
||||
"""Clean up resources."""
|
||||
self._stop_event.set()
|
||||
self.stop_watching()
|
||||
self._on_change_callbacks = []
|
||||
self._locator.stop()
|
||||
self._started = False
|
||||
|
||||
|
||||
# Singleton instance
|
||||
_leader_client: Optional[LeaderClient] = None
|
||||
|
||||
|
||||
def get_leader_client() -> Optional[LeaderClient]:
|
||||
"""Get or create the LeaderClient singleton.
|
||||
|
||||
Returns None if META_CORE_PATH is not set.
|
||||
"""
|
||||
"""Process-wide singleton, so one UDP socket serves every caller."""
|
||||
global _leader_client
|
||||
if _leader_client is None:
|
||||
meta_core_path = os.environ.get('META_CORE_PATH', '/meta-core')
|
||||
meta_core_url = os.environ.get('META_CORE_URL')
|
||||
_leader_client = LeaderClient(meta_core_path, meta_core_url)
|
||||
_leader_client = LeaderClient(
|
||||
meta_core_path=os.environ.get("META_CORE_PATH"),
|
||||
meta_core_url=os.environ.get("META_CORE_URL"),
|
||||
base_url=os.environ.get("BASE_URL") or None,
|
||||
)
|
||||
return _leader_client
|
||||
|
||||
Reference in New Issue
Block a user