移除 PolyWeather 市场提醒功能及监控基础设施

- 删除 market_alert_engine.py:交易预警引擎
- 删除 alertmanager_telegram_relay.py:Alertmanager 到 Telegram 转发
- 移除 telegram_push.py 中市场监控推送循环
- 移除 runtime_coordinator.py 中 trade_alert_push 协程
- 移除 docker-compose.yml 中 Prometheus/Alertmanager/Grafana 服务
- 移除 monitoring/ 目录:prometheus/alertmanager/grafana 配置

Tested: ruff check . ✓
This commit is contained in:
2569718930@qq.com
2026-05-14 18:23:39 +08:00
parent 5ad89b4c29
commit eb056a890b
11 changed files with 2 additions and 2069 deletions
-65
View File
@@ -45,68 +45,3 @@ services:
- ${POLYWEATHER_RUNTIME_DATA_DIR:-/var/lib/polyweather}:/app/data
user: "${UID:-1000}:${GID:-1000}"
polyweather_prometheus:
image: prom/prometheus:v3.4.1
container_name: polyweather_prometheus
restart: unless-stopped
profiles: ["monitoring"]
depends_on:
- polyweather_web
command:
- "--config.file=/etc/prometheus/prometheus.yml"
- "--storage.tsdb.path=/prometheus"
- "--storage.tsdb.retention.time=15d"
- "--web.enable-lifecycle"
volumes:
- ./monitoring/prometheus/prometheus.yml:/etc/prometheus/prometheus.yml:ro
- ./monitoring/prometheus/alerts.yml:/etc/prometheus/alerts.yml:ro
- ${POLYWEATHER_RUNTIME_DATA_DIR:-/var/lib/polyweather}/monitoring/prometheus:/prometheus
ports:
- "${POLYWEATHER_PROMETHEUS_PORT:-9090}:9090"
polyweather_alertmanager:
image: prom/alertmanager:v0.28.1
container_name: polyweather_alertmanager
restart: unless-stopped
profiles: ["monitoring"]
depends_on:
- polyweather_alert_relay
command:
- "--config.file=/etc/alertmanager/alertmanager.yml"
- "--storage.path=/alertmanager"
volumes:
- ./monitoring/alertmanager/alertmanager.yml:/etc/alertmanager/alertmanager.yml:ro
- ${POLYWEATHER_RUNTIME_DATA_DIR:-/var/lib/polyweather}/monitoring/alertmanager:/alertmanager
ports:
- "${POLYWEATHER_ALERTMANAGER_PORT:-9093}:9093"
polyweather_alert_relay:
<<: *polyweather-base
container_name: polyweather_alert_relay
restart: unless-stopped
profiles: ["monitoring"]
command: python scripts/alertmanager_telegram_relay.py
volumes:
- ${POLYWEATHER_RUNTIME_DATA_DIR:-/var/lib/polyweather}:/var/lib/polyweather
- ${POLYWEATHER_RUNTIME_DATA_DIR:-/var/lib/polyweather}:/app/data
ports:
- "${POLYWEATHER_ALERT_RELAY_PORT:-9099}:9099"
user: "${UID:-1000}:${GID:-1000}"
polyweather_grafana:
image: grafana/grafana-oss:12.0.2
container_name: polyweather_grafana
restart: unless-stopped
profiles: ["monitoring"]
depends_on:
- polyweather_prometheus
environment:
GF_SECURITY_ADMIN_USER: ${POLYWEATHER_GRAFANA_ADMIN_USER:-admin}
GF_SECURITY_ADMIN_PASSWORD: ${POLYWEATHER_GRAFANA_ADMIN_PASSWORD:-polyweather}
GF_USERS_ALLOW_SIGN_UP: "false"
volumes:
- ./monitoring/grafana/provisioning:/etc/grafana/provisioning:ro
- ./monitoring/grafana/dashboards:/var/lib/grafana/dashboards:ro
- ${POLYWEATHER_RUNTIME_DATA_DIR:-/var/lib/polyweather}/monitoring/grafana:/var/lib/grafana
ports:
- "${POLYWEATHER_GRAFANA_PORT:-3001}:3000"
-15
View File
@@ -1,15 +0,0 @@
global:
resolve_timeout: 5m
route:
receiver: telegram-relay
group_by: ["alertname"]
group_wait: 30s
group_interval: 5m
repeat_interval: 3h
receivers:
- name: telegram-relay
webhook_configs:
- url: http://polyweather_alert_relay:9099/alerts
send_resolved: true
@@ -1,227 +0,0 @@
{
"annotations": {
"list": [
{
"builtIn": 1,
"datasource": {
"type": "grafana",
"uid": "-- Grafana --"
},
"enable": true,
"hide": true,
"iconColor": "rgba(0, 211, 255, 1)",
"name": "Annotations & Alerts",
"type": "dashboard"
}
]
},
"editable": true,
"fiscalYearStartMonth": 0,
"graphTooltip": 0,
"id": null,
"links": [],
"panels": [
{
"datasource": {
"type": "prometheus",
"uid": "prometheus"
},
"fieldConfig": {
"defaults": {
"color": {
"mode": "palette-classic"
},
"unit": "reqps"
},
"overrides": []
},
"gridPos": {
"h": 8,
"w": 12,
"x": 0,
"y": 0
},
"id": 1,
"options": {
"legend": {
"displayMode": "table",
"placement": "bottom",
"showLegend": true
},
"tooltip": {
"mode": "single"
}
},
"targets": [
{
"expr": "sum by (status) (rate(polyweather_http_requests_total[5m]))",
"legendFormat": "{{status}}",
"refId": "A"
}
],
"title": "HTTP Requests by Status",
"type": "timeseries"
},
{
"datasource": {
"type": "prometheus",
"uid": "prometheus"
},
"fieldConfig": {
"defaults": {
"color": {
"mode": "palette-classic"
},
"unit": "ms"
},
"overrides": []
},
"gridPos": {
"h": 8,
"w": 12,
"x": 12,
"y": 0
},
"id": 2,
"options": {
"legend": {
"displayMode": "table",
"placement": "bottom",
"showLegend": true
},
"tooltip": {
"mode": "single"
}
},
"targets": [
{
"expr": "sum(rate(polyweather_http_request_duration_ms_sum[5m])) / clamp_min(sum(rate(polyweather_http_request_duration_ms_count[5m])), 1)",
"legendFormat": "avg",
"refId": "A"
},
{
"expr": "max(polyweather_http_request_duration_ms_max)",
"legendFormat": "max",
"refId": "B"
}
],
"title": "HTTP Latency",
"type": "timeseries"
},
{
"datasource": {
"type": "prometheus",
"uid": "prometheus"
},
"fieldConfig": {
"defaults": {
"color": {
"mode": "palette-classic"
},
"unit": "reqps"
},
"overrides": []
},
"gridPos": {
"h": 8,
"w": 12,
"x": 0,
"y": 8
},
"id": 3,
"options": {
"legend": {
"displayMode": "table",
"placement": "bottom",
"showLegend": true
},
"tooltip": {
"mode": "single"
}
},
"targets": [
{
"expr": "sum by (source, outcome) (rate(polyweather_source_requests_total[5m]))",
"legendFormat": "{{source}} / {{outcome}}",
"refId": "A"
}
],
"title": "Source Requests by Outcome",
"type": "timeseries"
},
{
"datasource": {
"type": "prometheus",
"uid": "prometheus"
},
"fieldConfig": {
"defaults": {
"color": {
"mode": "thresholds"
},
"thresholds": {
"mode": "absolute",
"steps": [
{
"color": "green"
},
{
"color": "orange",
"value": 10
},
{
"color": "red",
"value": 25
}
]
},
"unit": "percentunit"
},
"overrides": []
},
"gridPos": {
"h": 8,
"w": 12,
"x": 12,
"y": 8
},
"id": 4,
"options": {
"reduceOptions": {
"calcs": [
"lastNotNull"
],
"fields": "",
"values": false
}
},
"targets": [
{
"expr": "(sum(increase(polyweather_source_requests_total{outcome!~\"success|cache_hit\"}[15m])) / clamp_min(sum(increase(polyweather_source_requests_total[15m])), 1))",
"refId": "A"
}
],
"title": "Source Error Rate (15m)",
"type": "stat"
}
],
"refresh": "30s",
"schemaVersion": 41,
"style": "dark",
"tags": [
"polyweather",
"ops"
],
"templating": {
"list": []
},
"time": {
"from": "now-6h",
"to": "now"
},
"timepicker": {},
"timezone": "browser",
"title": "PolyWeather Overview",
"uid": "polyweather-overview",
"version": 1
}
@@ -1,11 +0,0 @@
apiVersion: 1
providers:
- name: PolyWeather Dashboards
orgId: 1
folder: PolyWeather
type: file
disableDeletion: false
editable: true
options:
path: /var/lib/grafana/dashboards
@@ -1,9 +0,0 @@
apiVersion: 1
datasources:
- name: PolyWeather Prometheus
type: prometheus
access: proxy
url: http://polyweather_prometheus:9090
isDefault: true
editable: false
-58
View File
@@ -1,58 +0,0 @@
groups:
- name: polyweather-runtime
rules:
- alert: PolyWeatherWebDown
expr: up{job="polyweather-web"} == 0
for: 2m
labels:
severity: critical
annotations:
summary: "PolyWeather web metrics endpoint is down"
description: "/metrics on polyweather_web has been unreachable for more than 2 minutes."
- alert: PolyWeatherHttp5xxBurst
expr: sum(increase(polyweather_http_requests_total{status=~"5.."}[10m])) > 10
for: 5m
labels:
severity: warning
annotations:
summary: "PolyWeather HTTP 5xx burst"
description: "More than 10 server errors were observed in the last 10 minutes."
- alert: PolyWeatherHighSourceErrorRate
expr: |
(
sum(increase(polyweather_source_requests_total{outcome!~"success|cache_hit"}[15m]))
/
clamp_min(sum(increase(polyweather_source_requests_total[15m])), 1)
) > 0.25
and sum(increase(polyweather_source_requests_total[15m])) > 20
for: 10m
labels:
severity: warning
annotations:
summary: "PolyWeather source error rate is high"
description: "External weather source errors exceeded 25% over the last 15 minutes."
- alert: PolyWeatherOpenMeteoCooldownLoop
expr: sum(increase(polyweather_source_requests_total{source="open_meteo",outcome=~"cooldown_skip|error"}[15m])) > 20
for: 10m
labels:
severity: warning
annotations:
summary: "Open-Meteo is rate-limited or erroring"
description: "Open-Meteo has produced repeated cooldown skips or errors in the last 15 minutes."
- alert: PolyWeatherSlowHttpAverage
expr: |
(
sum(rate(polyweather_http_request_duration_ms_sum[10m]))
/
clamp_min(sum(rate(polyweather_http_request_duration_ms_count[10m])), 1)
) > 2000
for: 10m
labels:
severity: warning
annotations:
summary: "PolyWeather average HTTP latency is high"
description: "Average HTTP request duration exceeded 2s over the last 10 minutes."
-19
View File
@@ -1,19 +0,0 @@
global:
scrape_interval: 30s
evaluation_interval: 30s
rule_files:
- /etc/prometheus/alerts.yml
alerting:
alertmanagers:
- static_configs:
- targets:
- polyweather_alertmanager:9093
scrape_configs:
- job_name: polyweather-web
metrics_path: /metrics
static_configs:
- targets:
- polyweather_web:8000
-107
View File
@@ -1,107 +0,0 @@
from __future__ import annotations
import json
import os
from http.server import BaseHTTPRequestHandler, HTTPServer
from typing import Any, Dict, List
import requests
def _chat_ids() -> List[str]:
raw = (
os.getenv("POLYWEATHER_MONITORING_ALERT_CHAT_IDS")
or os.getenv("TELEGRAM_CHAT_IDS")
or os.getenv("TELEGRAM_CHAT_ID")
or ""
)
return [item.strip() for item in raw.split(",") if item.strip()]
def _format_alerts(payload: Dict[str, Any]) -> str:
alerts = payload.get("alerts") or []
if not isinstance(alerts, list) or not alerts:
return "PolyWeather monitoring received an empty alert payload."
lines = ["PolyWeather monitoring alert"]
for alert in alerts[:10]:
if not isinstance(alert, dict):
continue
status = str(alert.get("status") or "unknown").upper()
labels = alert.get("labels") or {}
annotations = alert.get("annotations") or {}
alert_name = labels.get("alertname") or "unknown_alert"
severity = labels.get("severity") or "info"
summary = annotations.get("summary") or annotations.get("description") or ""
lines.append(f"- [{status}] {alert_name} ({severity})")
if summary:
lines.append(f" {summary}")
return "\n".join(lines)
def _send_telegram_message(text: str) -> None:
token = str(os.getenv("TELEGRAM_BOT_TOKEN") or "").strip()
chat_ids = _chat_ids()
if not token or not chat_ids:
return
url = f"https://api.telegram.org/bot{token}/sendMessage"
for chat_id in chat_ids:
try:
requests.post(
url,
json={
"chat_id": chat_id,
"text": text,
"disable_web_page_preview": True,
},
timeout=10,
).raise_for_status()
except Exception:
continue
def _send_alert_notifications(text: str) -> None:
_send_telegram_message(text)
class _Handler(BaseHTTPRequestHandler):
def do_GET(self) -> None: # noqa: N802
if self.path.rstrip("/") == "/healthz":
self.send_response(200)
self.send_header("Content-Type", "application/json; charset=utf-8")
self.end_headers()
self.wfile.write(b'{"ok":true}')
return
self.send_error(404)
def do_POST(self) -> None: # noqa: N802
if self.path.rstrip("/") != "/alerts":
self.send_error(404)
return
length = int(self.headers.get("Content-Length", "0") or "0")
raw = self.rfile.read(length) if length > 0 else b"{}"
try:
payload = json.loads(raw.decode("utf-8"))
except Exception:
self.send_error(400, "invalid json")
return
if not isinstance(payload, dict):
self.send_error(400, "invalid payload")
return
_send_alert_notifications(_format_alerts(payload))
self.send_response(200)
self.send_header("Content-Type", "application/json; charset=utf-8")
self.end_headers()
self.wfile.write(b'{"ok":true}')
def log_message(self, format: str, *args: object) -> None: # noqa: A003
return
def main() -> None:
port = 9099
server = HTTPServer(("0.0.0.0", port), _Handler)
server.serve_forever()
if __name__ == "__main__":
main()
File diff suppressed because it is too large Load Diff
+2 -35
View File
@@ -83,7 +83,6 @@ class StartupCoordinator:
def start_all(self) -> RuntimeStatus:
loops = [
self._start_trade_alert_loop(),
self._start_airport_high_freq_loop(),
self._start_dashboard_prewarm_loop(),
self._start_polygon_wallet_loop(),
@@ -152,37 +151,6 @@ class StartupCoordinator:
details=details,
)
def _start_trade_alert_loop(self) -> LoopStatus:
enabled = _env_bool("TELEGRAM_ALERT_PUSH_ENABLED", True)
chat_ids = get_telegram_chat_ids_from_env()
interval = max(60, _env_int("TELEGRAM_ALERT_PUSH_INTERVAL_SEC", 1800))
cities_count = _parse_csv_count(os.getenv("TELEGRAM_ALERT_CITIES"))
details = {
"mode": "critical-alerts",
"interval_sec": interval,
"cities_count": cities_count,
"chat_targets": len(chat_ids),
"alert_cooldown_sec": max(
60,
_env_int("TELEGRAM_ALERT_PUSH_COOLDOWN_SEC", 21600),
),
"min_trigger_count": max(1, _env_int("TELEGRAM_ALERT_MIN_TRIGGER_COUNT", 3)),
"min_severity": str(os.getenv("TELEGRAM_ALERT_MIN_SEVERITY") or "high").strip().lower(),
"daily_signal_cap": "none",
}
validation_error = None if chat_ids else "missing_TELEGRAM_CHAT_IDS"
return self._start_with_validation(
key="trade_alert_push",
label="市场监控推送",
configured_enabled=enabled,
details=details,
validation_error=validation_error,
starter=lambda: import_module("src.utils.telegram_push").start_trade_alert_push_loop(
self.bot,
self.config,
),
)
def _start_airport_high_freq_loop(self) -> LoopStatus:
enabled = _env_bool("TELEGRAM_AIRPORT_PUSH_ENABLED", True)
chat_ids = get_telegram_chat_ids_from_env()
@@ -269,10 +237,9 @@ class StartupCoordinator:
label="Polymarket 钱包异动监听(已停用)",
configured_enabled=False,
started=False,
reason="retired_replaced_by_market_monitor",
reason="retired",
details={
"replacement": "trade_alert_push",
"note": "wallet activity watcher retired in favor of market monitor digests and critical alerts",
"note": "wallet activity watcher retired",
},
)
-205
View File
@@ -483,211 +483,6 @@ def _alert_signature(alert_payload: Dict[str, Any]) -> str:
raw = json.dumps(signature_payload, sort_keys=True, ensure_ascii=True)
return hashlib.sha1(raw.encode("utf-8")).hexdigest()
def build_trade_alert_for_city(
city: str,
config: Dict[str, Any],
force_refresh: bool = False,
target_date: Optional[str] = None,
) -> Dict[str, Any]:
from web.app import _analyze, _build_city_detail_payload
from src.analysis.market_alert_engine import build_trading_alerts
city_weather = _analyze(city, force_refresh=force_refresh)
try:
aggregate_detail = _build_city_detail_payload(
city_weather,
target_date=target_date,
)
market_scan = aggregate_detail.get("market_scan")
if isinstance(market_scan, dict):
city_weather = {**city_weather, "market_scan": market_scan}
except Exception as exc:
logger.debug(f"market scan attach skipped city={city}: {exc}")
resolved_target_date = target_date or city_weather.get("local_date")
if resolved_target_date:
datetime.strptime(resolved_target_date, "%Y-%m-%d")
map_url = os.getenv("POLYWEATHER_MAP_URL") or "https://polyweather-pro.vercel.app/"
alert_payload = build_trading_alerts(
city_weather=city_weather,
map_url=map_url,
)
alert_payload["target_date"] = resolved_target_date
return alert_payload
def _maybe_send_alert(
bot: Any,
chat_ids: List[str],
city: str,
alert_payload: Dict[str, Any],
state: Dict[str, Any],
cooldown_sec: int,
min_severity: str,
min_trigger_count: int,
) -> bool:
now_ts = int(time.time())
last_by_city = state.setdefault("last_by_city", {})
last_city = last_by_city.get(city) or {}
is_active = _severity_ok(alert_payload, min_severity, min_trigger_count)
if not _market_price_cap_ok(alert_payload):
is_active = False
message = ((alert_payload.get("telegram") or {}).get("zh") or "").strip()
if not is_active or not message:
if last_city.get("active"):
last_by_city[city] = {
**last_city,
"active": False,
"cleared_ts": now_ts,
}
logger.info(f"market monitor disarmed city={city}")
return True
return False
if not chat_ids:
return False
signature = _alert_signature(alert_payload)
trigger_key = _trigger_type_key(alert_payload)
last_city_sig = last_city.get("signature")
last_city_key = str(last_city.get("trigger_key") or "")
last_city_ts = int(last_city.get("ts") or 0)
last_sig_ts = int((state.get("by_signature") or {}).get(signature) or 0)
last_city_active = bool(last_city.get("active"))
if last_city_active and last_city_key == trigger_key and last_city_sig == signature:
return False
if last_city_ts and now_ts - last_city_ts < cooldown_sec:
return False
if last_sig_ts and now_ts - last_sig_ts < cooldown_sec:
return False
sent_count = 0
for chat_id in chat_ids:
try:
bot.send_message(chat_id, message)
sent_count += 1
except Exception as exc:
logger.warning("market monitor push failed city={} chat_id={} error={}", city, chat_id, exc)
if sent_count <= 0:
return False
last_by_city[city] = {
"signature": signature,
"trigger_key": trigger_key,
"severity": alert_payload.get("severity"),
"ts": now_ts,
"active": True,
"evidence": alert_payload.get("evidence"),
}
state.setdefault("by_signature", {})[signature] = now_ts
logger.info(
f"market monitor pushed city={city} severity={alert_payload.get('severity')} "
f"trigger_count={alert_payload.get('trigger_count')} trigger_key={trigger_key} "
f"evidence={_evidence_brief(alert_payload)} chat_targets={sent_count}"
)
return True
def _run_market_monitor_cycle(
bot: Any,
config: Dict[str, Any],
*,
chat_ids: List[str],
cities: List[str],
state: Dict[str, Any],
alert_cooldown_sec: int,
min_severity: str,
min_trigger_count: int,
sleep_between_cities_sec: float = 1.0,
) -> bool:
state_dirty = False
for city in cities:
try:
alert_payload = build_trade_alert_for_city(city, config)
if _maybe_send_alert(
bot=bot,
chat_ids=chat_ids,
city=city,
alert_payload=alert_payload,
state=state,
cooldown_sec=alert_cooldown_sec,
min_severity=min_severity,
min_trigger_count=min_trigger_count,
):
state_dirty = True
except Exception:
logger.exception(f"telegram market monitor loop failed for city={city}")
if sleep_between_cities_sec > 0:
time.sleep(sleep_between_cities_sec)
return state_dirty
def start_trade_alert_push_loop(bot: Any, config: Dict[str, Any]) -> Optional[threading.Thread]:
enabled = _env_bool("TELEGRAM_ALERT_PUSH_ENABLED", True)
chat_ids = get_telegram_chat_ids_from_env()
if not enabled:
logger.info("telegram market monitor loop disabled")
return None
if not chat_ids:
logger.warning("telegram market monitor loop skipped: TELEGRAM_CHAT_IDS is not set")
return None
interval_sec = max(60, _env_int("TELEGRAM_ALERT_PUSH_INTERVAL_SEC", 1800))
cities = _parse_city_list(os.getenv("TELEGRAM_ALERT_CITIES"))
state_path = _state_file()
alert_cooldown_sec = max(60, _env_int("TELEGRAM_ALERT_PUSH_COOLDOWN_SEC", 21600))
min_trigger_count = max(1, _env_int("TELEGRAM_ALERT_MIN_TRIGGER_COUNT", 3))
min_severity = str(os.getenv("TELEGRAM_ALERT_MIN_SEVERITY") or "high").strip().lower()
if min_severity not in SEVERITY_RANK:
min_severity = "high"
def _runner() -> None:
try:
_save_state(state_path, _load_state(state_path))
except Exception:
logger.exception(f"failed to initialize market monitor state path={state_path}")
logger.info(
f"telegram market monitor loop started mode=critical-alerts "
f"cities={len(cities)} interval={interval_sec}s chat_targets={len(chat_ids)} "
f"alert_cooldown={alert_cooldown_sec}s "
f"min_severity={min_severity} min_trigger_count={min_trigger_count} "
f"state_path={state_path}"
)
while True:
cycle_started = time.time()
state = _load_state(state_path)
_cleanup_state(state, int(cycle_started))
if _run_market_monitor_cycle(
bot=bot,
config=config,
chat_ids=chat_ids,
cities=cities,
state=state,
alert_cooldown_sec=alert_cooldown_sec,
min_severity=min_severity,
min_trigger_count=min_trigger_count,
):
_save_state(state_path, state)
elapsed = time.time() - cycle_started
sleep_sec = max(5, interval_sec - int(elapsed))
time.sleep(sleep_sec)
thread = threading.Thread(
target=_runner,
name="telegram-market-monitor-pusher",
daemon=True,
)
thread.start()
return thread
# ── high-freq airport push loop ──
HIGH_FREQ_AIRPORT_CITIES = {"seoul", "busan", "tokyo", "ankara", "helsinki", "amsterdam", "istanbul", "paris", "hong kong", "lau fau shan", "taipei"}