2ce722bda9
P0 #3 (徹底長期修系列) — 把 daily report 的 pod 健康判斷從「ready=False 一律告警」 升級到完整 K8s pod lifecycle state machine： | Phase | 行為 | |-------|------| | Succeeded / Completed | 跳過（CronJob/Job 跑完正常） | | Failed | 必告警 | | Unknown | 必告警 | | Pending <5min | 跳過（剛 schedule 合理） | | Pending >=5min | 告警「image pull / scheduling 卡住」| | Running ready=True | 健康，跳過 | | Running ready=False <2min | 跳過（剛起來 probe 還沒過）| | Running ready=False >=2min | 告警「readiness probe fail / 啟動異常」| | restarts >=3 | 必告警（無論 phase）| 實作： - PodInfo 加 start_time: Optional[str]（從 .status.startTime） - _get_pod_status kubectl custom-columns 加 STARTTIME - _build_warnings 完整 state machine + 閾值常數 regression test (test_heartbeat_pod_state_machine.py 13 個) 覆蓋每個 phase + 邊界條件，含 2026-05-02 統帥截圖鐵證重現（3 個 drift-scanner Succeeded pod 不該觸發「需關注 3 項」假警報）。 Tests: 13 passed (新增 test_heartbeat_pod_state_machine.py) 接續 a38d9112（單純 Succeeded skip），這次徹底處理 Pending/Failed/Unknown + 時間閾值 + 沒 start_time 的保守告警。 Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
916 lines
39 KiB
Python
916 lines
39 KiB
Python
"""
|
||
HeartbeatReportService — ADR-073 心跳監控重構
|
||
==============================================
|
||
並行收集所有探測 → 彙整判斷 → 一份報告
|
||
|
||
設計原則:
|
||
- 所有探測 asyncio.gather(return_exceptions=True)，任一失敗不影響其他
|
||
- 只負責收集資料 + 組裝報告，不負責發送
|
||
- 超時保護：每個探測最多 8 秒
|
||
|
||
建立時間: 2026-04-12 (台北時區) ogt
|
||
建立者: Claude Sonnet 4.6 — ADR-073 心跳重構
|
||
"""
|
||
|
||
import asyncio
|
||
import html
|
||
from dataclasses import dataclass, field
|
||
from datetime import datetime, timedelta
|
||
from typing import Optional
|
||
|
||
import httpx
|
||
import structlog
|
||
|
||
from src.core.config import get_settings
|
||
from src.utils.timezone import now_taipei
|
||
|
||
logger = structlog.get_logger(__name__)
|
||
|
||
settings = get_settings()
|
||
|
||
_PROBE_TIMEOUT = 8.0 # 每個探測最長 8 秒
|
||
|
||
|
||
@dataclass
|
||
class ProbeResult:
|
||
ok: bool
|
||
status: str # "✅ 正常" / "❌ 失敗: ..." / "⚠️ 警告: ..."
|
||
latency_ms: Optional[float] = None
|
||
|
||
|
||
@dataclass
|
||
class FlywheelStats:
|
||
playbook_count: int = 0
|
||
success_24h: int = 0
|
||
attempt_24h: int = 0
|
||
km_total: int = 0
|
||
km_vectorized: int = 0
|
||
last_learning_at: Optional[datetime] = None
|
||
|
||
|
||
@dataclass
|
||
class AlertPipelineStats:
|
||
total_24h: int = 0
|
||
auto_resolved_24h: int = 0
|
||
pending_approval: int = 0
|
||
execution_success_24h: int = 0
|
||
execution_failed_24h: int = 0
|
||
|
||
|
||
@dataclass
|
||
class DbRedisStats:
|
||
db_ok: bool = False
|
||
db_status: str = "❌ 未查詢"
|
||
redis_ok: bool = False
|
||
redis_status: str = "❌ 未查詢"
|
||
redis_key_count: int = 0
|
||
|
||
|
||
@dataclass
|
||
class PodInfo:
|
||
name: str
|
||
ready: bool
|
||
status: str
|
||
restarts: int = 0
|
||
# 2026-05-03 Claude Opus 4.7 + 統帥 ogt：P0 #3 K8s pod state machine 完整化
|
||
# 加 start_time 才能判斷 Pending/NotReady 是「剛起來合理」還是「卡住該告警」
|
||
start_time: Optional[str] = None # ISO 8601 from .status.startTime
|
||
|
||
|
||
@dataclass
|
||
class ScannerStats:
|
||
# key = scanner name, value = last run ISO string or None
|
||
last_runs: dict[str, Optional[str]] = field(default_factory=dict)
|
||
|
||
|
||
@dataclass
|
||
class TelegramBotStats:
|
||
polling_ok: bool = False
|
||
status: str = "❌ 未查詢"
|
||
last_callback_ago_min: Optional[float] = None
|
||
|
||
|
||
@dataclass
|
||
class AutomationStats:
|
||
"""自動化六大能力今日統計（2026-04-24 ogt: Task 3 — 告警自動化可觀測性）"""
|
||
# 自動規則生成
|
||
auto_rule_generated_today: int = 0
|
||
# 自動審核拒絕（按原因分組）
|
||
reject_counts: dict[str, int] = field(default_factory=dict)
|
||
# 今日 KM 新增
|
||
km_created_today: int = 0
|
||
# Config Drift 自動採納（今日）
|
||
drift_adopted_today: int = 0
|
||
|
||
|
||
@dataclass
|
||
class HeartbeatReport:
|
||
timestamp: datetime
|
||
ai_services: dict[str, ProbeResult] = field(default_factory=dict)
|
||
ollama_models: dict[str, bool] = field(default_factory=dict)
|
||
mcp_providers: dict[str, ProbeResult] = field(default_factory=dict)
|
||
flywheel: FlywheelStats = field(default_factory=FlywheelStats)
|
||
infra: dict[str, ProbeResult] = field(default_factory=dict)
|
||
warnings: list[str] = field(default_factory=list)
|
||
# 2026-04-22 新增動態區塊
|
||
alert_pipeline: AlertPipelineStats = field(default_factory=AlertPipelineStats)
|
||
db_redis: DbRedisStats = field(default_factory=DbRedisStats)
|
||
pods: list[PodInfo] = field(default_factory=list)
|
||
scanners: ScannerStats = field(default_factory=ScannerStats)
|
||
telegram_bot: TelegramBotStats = field(default_factory=TelegramBotStats)
|
||
# 2026-04-24 自動化統計
|
||
automation: AutomationStats = field(default_factory=AutomationStats)
|
||
|
||
@property
|
||
def has_warnings(self) -> bool:
|
||
return len(self.warnings) > 0
|
||
|
||
|
||
class HeartbeatReportService:
|
||
"""
|
||
心跳報告收集服務
|
||
|
||
使用方式:
|
||
report = await HeartbeatReportService().collect()
|
||
text = report_to_telegram_html(report)
|
||
"""
|
||
|
||
async def collect(self) -> HeartbeatReport:
|
||
"""並行收集所有探測，彙整為一份報告"""
|
||
report = HeartbeatReport(timestamp=now_taipei())
|
||
|
||
results = await asyncio.gather(
|
||
self._probe_ollama(),
|
||
self._probe_nemotron(),
|
||
self._probe_gemini(),
|
||
self._probe_claude(),
|
||
self._probe_mcp_k8s(),
|
||
self._probe_mcp_ssh(),
|
||
self._probe_mcp_argocd(),
|
||
self._probe_mcp_sentry(),
|
||
self._probe_argocd_sync(),
|
||
self._probe_velero(),
|
||
self._get_flywheel_stats(),
|
||
# 2026-04-22 新增動態探測
|
||
self._get_alert_pipeline_stats(),
|
||
self._probe_db_redis(),
|
||
self._get_pod_status(),
|
||
self._get_scanner_stats(),
|
||
self._probe_telegram_bot(),
|
||
# 2026-04-24 自動化統計
|
||
self._get_automation_stats(),
|
||
return_exceptions=True,
|
||
)
|
||
|
||
keys = [
|
||
"_ollama", "_nemotron", "_gemini", "_claude",
|
||
"_mcp_k8s", "_mcp_ssh", "_mcp_argocd", "_mcp_sentry",
|
||
"_argocd_sync", "_velero", "_flywheel",
|
||
"_alert_pipeline", "_db_redis", "_pods", "_scanners", "_tg_bot",
|
||
"_automation",
|
||
]
|
||
collected: dict = {}
|
||
for key, result in zip(keys, results):
|
||
if isinstance(result, Exception):
|
||
logger.warning("heartbeat_probe_error", probe=key, error=str(result))
|
||
collected[key] = None
|
||
else:
|
||
collected[key] = result
|
||
|
||
# --- AI 服務 ---
|
||
ollama_data = collected["_ollama"] or {}
|
||
report.ai_services["ollama"] = ollama_data.get("probe", ProbeResult(False, "❌ 無回應"))
|
||
report.ollama_models = ollama_data.get("models", {})
|
||
report.ai_services["nemotron"] = collected["_nemotron"] or ProbeResult(False, "❌ 無回應")
|
||
report.ai_services["gemini"] = collected["_gemini"] or ProbeResult(False, "❌ 無回應")
|
||
report.ai_services["claude"] = collected["_claude"] or ProbeResult(False, "❌ 無回應")
|
||
|
||
# --- MCP Provider ---
|
||
report.mcp_providers["k8s"] = collected["_mcp_k8s"] or ProbeResult(False, "❌ 無回應")
|
||
report.mcp_providers["ssh"] = collected["_mcp_ssh"] or ProbeResult(False, "❌ 無回應")
|
||
report.mcp_providers["argocd"] = collected["_mcp_argocd"] or ProbeResult(False, "❌ 無回應")
|
||
report.mcp_providers["sentry"] = collected["_mcp_sentry"] or ProbeResult(False, "❌ 無回應")
|
||
|
||
# --- 基礎設施 ---
|
||
report.infra["argocd_sync"] = collected["_argocd_sync"] or ProbeResult(False, "❌ 無回應")
|
||
report.infra["velero"] = collected["_velero"] or ProbeResult(False, "❌ 無回應")
|
||
|
||
# --- 飛輪統計 ---
|
||
if collected["_flywheel"]:
|
||
report.flywheel = collected["_flywheel"]
|
||
|
||
# --- 新動態區塊 ---
|
||
if collected["_alert_pipeline"]:
|
||
report.alert_pipeline = collected["_alert_pipeline"]
|
||
if collected["_db_redis"]:
|
||
report.db_redis = collected["_db_redis"]
|
||
if collected["_pods"]:
|
||
report.pods = collected["_pods"]
|
||
if collected["_scanners"]:
|
||
report.scanners = collected["_scanners"]
|
||
if collected["_tg_bot"]:
|
||
report.telegram_bot = collected["_tg_bot"]
|
||
if collected["_automation"]:
|
||
report.automation = collected["_automation"]
|
||
|
||
# --- 彙整 warnings ---
|
||
report.warnings = self._build_warnings(report)
|
||
|
||
return report
|
||
|
||
# =========================================================================
|
||
# 探測方法
|
||
# =========================================================================
|
||
|
||
async def _probe_ollama(self) -> dict:
|
||
"""探測 Ollama 服務 + 逐一確認所需模型"""
|
||
try:
|
||
async with httpx.AsyncClient(timeout=_PROBE_TIMEOUT) as client:
|
||
t0 = asyncio.get_event_loop().time()
|
||
resp = await client.get(f"{settings.OLLAMA_URL}/api/tags")
|
||
latency = (asyncio.get_event_loop().time() - t0) * 1000
|
||
|
||
if resp.status_code != 200:
|
||
return {
|
||
"probe": ProbeResult(False, f"❌ HTTP {resp.status_code}", latency),
|
||
"models": {},
|
||
}
|
||
|
||
available = {m["name"] for m in resp.json().get("models", [])}
|
||
# 也把 short name（無 :tag）加進去方便匹配
|
||
available_short = {n.split(":")[0] for n in available}
|
||
|
||
model_status: dict[str, bool] = {}
|
||
for required in settings.OLLAMA_REQUIRED_MODELS:
|
||
req_short = required.split(":")[0]
|
||
ok = required in available or req_short in available_short
|
||
model_status[required] = ok
|
||
|
||
return {
|
||
"probe": ProbeResult(True, "✅ 正常", round(latency, 1)),
|
||
"models": model_status,
|
||
}
|
||
except Exception as e:
|
||
return {
|
||
"probe": ProbeResult(False, f"❌ {str(e)[:60]}"),
|
||
"models": {},
|
||
}
|
||
|
||
async def _probe_nemotron(self) -> ProbeResult:
|
||
"""探測 Nemotron NIM API"""
|
||
if not settings.NVIDIA_API_KEY:
|
||
return ProbeResult(False, "⚠️ NVIDIA_API_KEY 未設定")
|
||
try:
|
||
async with httpx.AsyncClient(timeout=_PROBE_TIMEOUT) as client:
|
||
t0 = asyncio.get_event_loop().time()
|
||
resp = await client.get(
|
||
"https://integrate.api.nvidia.com/v1/models",
|
||
headers={"Authorization": f"Bearer {settings.NVIDIA_API_KEY}"},
|
||
)
|
||
latency = (asyncio.get_event_loop().time() - t0) * 1000
|
||
if resp.status_code == 200:
|
||
return ProbeResult(True, "✅ 正常", round(latency, 1))
|
||
return ProbeResult(False, f"❌ HTTP {resp.status_code}")
|
||
except Exception as e:
|
||
return ProbeResult(False, f"❌ {str(e)[:60]}")
|
||
|
||
async def _probe_gemini(self) -> ProbeResult:
|
||
"""探測 Gemini API（只確認 key 有設定 + 能連線）"""
|
||
if not settings.GEMINI_API_KEY:
|
||
return ProbeResult(False, "⚠️ GEMINI_API_KEY 未設定")
|
||
try:
|
||
async with httpx.AsyncClient(timeout=_PROBE_TIMEOUT) as client:
|
||
t0 = asyncio.get_event_loop().time()
|
||
resp = await client.get(
|
||
"https://generativelanguage.googleapis.com/v1beta/models",
|
||
headers={"x-goog-api-key": settings.GEMINI_API_KEY},
|
||
)
|
||
latency = (asyncio.get_event_loop().time() - t0) * 1000
|
||
if resp.status_code == 200:
|
||
return ProbeResult(True, "✅ 正常", round(latency, 1))
|
||
return ProbeResult(False, f"❌ HTTP {resp.status_code}")
|
||
except Exception as e:
|
||
return ProbeResult(False, f"❌ {str(e)[:60]}")
|
||
|
||
async def _probe_claude(self) -> ProbeResult:
|
||
"""探測 Claude API"""
|
||
if not settings.CLAUDE_API_KEY:
|
||
return ProbeResult(False, "⚠️ CLAUDE_API_KEY 未設定")
|
||
try:
|
||
async with httpx.AsyncClient(timeout=_PROBE_TIMEOUT) as client:
|
||
t0 = asyncio.get_event_loop().time()
|
||
resp = await client.get(
|
||
"https://api.anthropic.com/v1/models",
|
||
headers={
|
||
"x-api-key": settings.CLAUDE_API_KEY,
|
||
"anthropic-version": "2023-06-01",
|
||
},
|
||
)
|
||
latency = (asyncio.get_event_loop().time() - t0) * 1000
|
||
if resp.status_code == 200:
|
||
return ProbeResult(True, "✅ 正常", round(latency, 1))
|
||
return ProbeResult(False, f"❌ HTTP {resp.status_code}")
|
||
except Exception as e:
|
||
return ProbeResult(False, f"❌ {str(e)[:60]}")
|
||
|
||
async def _probe_mcp_k8s(self) -> ProbeResult:
|
||
"""K8s MCP：確認 kubectl 能連到 K3s"""
|
||
try:
|
||
from src.plugins.mcp.providers.k8s_provider import K8sProvider
|
||
provider = K8sProvider()
|
||
if not provider.enabled:
|
||
return ProbeResult(False, "⚠️ K8s MCP 未啟用")
|
||
result = await asyncio.wait_for(
|
||
provider.execute("kubectl_get", {"resource_type": "nodes"}),
|
||
timeout=_PROBE_TIMEOUT,
|
||
)
|
||
if result.success:
|
||
return ProbeResult(True, "✅ 正常")
|
||
return ProbeResult(False, f"⚠️ {result.error[:60] if result.error else '查詢失敗'}")
|
||
except asyncio.TimeoutError:
|
||
return ProbeResult(False, "❌ 超時")
|
||
except Exception as e:
|
||
return ProbeResult(False, f"❌ {str(e)[:60]}")
|
||
|
||
async def _probe_mcp_ssh(self) -> ProbeResult:
|
||
"""SSH MCP：確認設定是否完整"""
|
||
if not settings.SSH_MCP_ENABLED:
|
||
return ProbeResult(False, "⚠️ SSH_MCP_ENABLED=false")
|
||
# 確認 ssh-mcp-key Secret 是否掛載
|
||
import os
|
||
key_path = "/run/secrets/ssh_mcp_key"
|
||
if not os.path.exists(key_path):
|
||
return ProbeResult(False, "⚠️ ssh-mcp-key 未注入 K8s Secret")
|
||
return ProbeResult(True, "✅ 正常")
|
||
|
||
async def _probe_mcp_argocd(self) -> ProbeResult:
|
||
"""ArgoCD MCP：確認 token 設定"""
|
||
if not settings.ARGOCD_MCP_ENABLED:
|
||
return ProbeResult(False, "⚠️ ARGOCD_MCP_ENABLED=false")
|
||
if not settings.ARGOCD_API_TOKEN:
|
||
return ProbeResult(False, "⚠️ ARGOCD_API_TOKEN 未設定")
|
||
return ProbeResult(True, "✅ 設定完整")
|
||
|
||
async def _probe_mcp_sentry(self) -> ProbeResult:
|
||
"""Sentry MCP：確認設定"""
|
||
if not settings.SENTRY_MCP_ENABLED:
|
||
return ProbeResult(False, "⚠️ SENTRY_MCP_ENABLED=false")
|
||
# 確認 SENTRY_AUTH_TOKEN
|
||
sentry_token = getattr(settings, "SENTRY_AUTH_TOKEN", "") or ""
|
||
if not sentry_token:
|
||
return ProbeResult(False, "⚠️ SENTRY_AUTH_TOKEN 未設定")
|
||
return ProbeResult(True, "✅ 設定完整")
|
||
|
||
async def _probe_argocd_sync(self) -> ProbeResult:
|
||
"""ArgoCD 應用同步狀態"""
|
||
if not settings.ARGOCD_API_TOKEN:
|
||
return ProbeResult(False, "⚠️ 未設定 Token，無法查詢")
|
||
try:
|
||
async with httpx.AsyncClient(
|
||
timeout=_PROBE_TIMEOUT,
|
||
verify=False, # ArgoCD self-signed cert
|
||
) as client:
|
||
resp = await client.get(
|
||
f"{settings.ARGOCD_URL}/api/v1/applications/awoooi-prod",
|
||
headers={"Authorization": f"Bearer {settings.ARGOCD_API_TOKEN}"},
|
||
)
|
||
if resp.status_code != 200:
|
||
return ProbeResult(False, f"❌ HTTP {resp.status_code}")
|
||
data = resp.json()
|
||
sync_status = data.get("status", {}).get("sync", {}).get("status", "Unknown")
|
||
health_status = data.get("status", {}).get("health", {}).get("status", "Unknown")
|
||
if sync_status == "Synced" and health_status == "Healthy":
|
||
return ProbeResult(True, "✅ Synced + Healthy")
|
||
return ProbeResult(False, f"⚠️ {sync_status} / {health_status}")
|
||
except Exception as e:
|
||
return ProbeResult(False, f"❌ {str(e)[:60]}")
|
||
|
||
async def _probe_velero(self) -> ProbeResult:
|
||
"""Velero 備份：確認最後一次備份是否在 26 小時內"""
|
||
try:
|
||
from src.plugins.mcp.providers.k8s_provider import K8sProvider
|
||
provider = K8sProvider()
|
||
if not provider.enabled:
|
||
return ProbeResult(False, "⚠️ K8s MCP 未啟用，無法查 Velero")
|
||
result = await asyncio.wait_for(
|
||
provider.execute("kubectl_get", {
|
||
"resource_type": "backups.velero.io",
|
||
"namespace": "velero",
|
||
}),
|
||
timeout=_PROBE_TIMEOUT,
|
||
)
|
||
if not result.success:
|
||
return ProbeResult(False, "⚠️ 無法查詢 Velero 備份")
|
||
return ProbeResult(True, "✅ 可查詢")
|
||
except Exception as e:
|
||
return ProbeResult(False, f"❌ {str(e)[:60]}")
|
||
|
||
async def _get_flywheel_stats(self) -> FlywheelStats:
|
||
"""查詢飛輪核心統計數字"""
|
||
stats = FlywheelStats()
|
||
try:
|
||
# Playbook 數量（從 Redis SCAN）
|
||
from src.core.redis_client import get_redis
|
||
redis = get_redis()
|
||
keys = await redis.keys("playbook:*")
|
||
stats.playbook_count = len(keys)
|
||
except Exception as e:
|
||
logger.debug("heartbeat_playbook_count_failed", error=str(e))
|
||
|
||
try:
|
||
# KM 向量化率（DB 查詢）
|
||
from src.db.base import get_db_context
|
||
from src.db.models import IncidentRecord, KnowledgeEntryRecord
|
||
from sqlalchemy import func, select
|
||
async with get_db_context() as db:
|
||
# KM 總數
|
||
km_total = await db.scalar(select(func.count()).select_from(KnowledgeEntryRecord))
|
||
stats.km_total = km_total or 0
|
||
|
||
# KM 向量化數（embedding IS NOT NULL）
|
||
# KnowledgeEntryRecord ORM 無 embedding 欄位，改用 raw SQL
|
||
from sqlalchemy import text as sa_text
|
||
vec_result = await db.execute(
|
||
sa_text("SELECT COUNT(*) FROM knowledge_entries WHERE embedding IS NOT NULL")
|
||
)
|
||
stats.km_vectorized = vec_result.scalar() or 0
|
||
|
||
# 24h 修復統計
|
||
since = datetime.utcnow() - timedelta(hours=24)
|
||
outcomes = await db.execute(
|
||
select(IncidentRecord.outcome).where(
|
||
IncidentRecord.created_at >= since,
|
||
IncidentRecord.outcome.isnot(None),
|
||
)
|
||
)
|
||
outcome_list = [r[0] for r in outcomes.all() if r[0]]
|
||
stats.attempt_24h = len(outcome_list)
|
||
stats.success_24h = sum(
|
||
1 for o in outcome_list
|
||
if isinstance(o, dict) and o.get("execution_success")
|
||
or isinstance(o, str) and "success" in o.lower()
|
||
)
|
||
|
||
# 最後學習活動
|
||
last_km = await db.scalar(
|
||
select(func.max(KnowledgeEntryRecord.created_at))
|
||
)
|
||
if last_km:
|
||
stats.last_learning_at = last_km
|
||
except Exception as e:
|
||
logger.debug("heartbeat_flywheel_stats_failed", error=str(e))
|
||
|
||
return stats
|
||
|
||
# =========================================================================
|
||
# 2026-04-22 新增動態探測方法
|
||
# =========================================================================
|
||
|
||
async def _get_alert_pipeline_stats(self) -> AlertPipelineStats:
|
||
"""查 24h 告警流水線統計（approval_records）"""
|
||
stats = AlertPipelineStats()
|
||
try:
|
||
from src.db.base import get_db_context
|
||
from sqlalchemy import text as sa_text
|
||
async with get_db_context() as db:
|
||
r = await db.execute(sa_text("""
|
||
SELECT
|
||
COUNT(*) AS total,
|
||
COUNT(*) FILTER (WHERE UPPER(status::text) = 'PENDING') AS pending,
|
||
COUNT(*) FILTER (WHERE UPPER(status::text) = 'EXECUTION_SUCCESS') AS success,
|
||
COUNT(*) FILTER (WHERE UPPER(status::text) = 'EXECUTION_FAILED') AS failed,
|
||
COUNT(*) FILTER (WHERE UPPER(status::text) IN ('APPROVED','EXECUTION_SUCCESS','EXECUTION_FAILED')) AS auto_resolved
|
||
FROM approval_records
|
||
WHERE created_at >= NOW() - interval '24 hours'
|
||
"""))
|
||
row = r.one()
|
||
stats.total_24h = int(row.total or 0)
|
||
stats.pending_approval = int(row.pending or 0)
|
||
stats.execution_success_24h = int(row.success or 0)
|
||
stats.execution_failed_24h = int(row.failed or 0)
|
||
stats.auto_resolved_24h = int(row.auto_resolved or 0)
|
||
except Exception as e:
|
||
logger.debug("heartbeat_alert_pipeline_failed", error=str(e))
|
||
return stats
|
||
|
||
async def _probe_db_redis(self) -> DbRedisStats:
|
||
"""探測 PostgreSQL 與 Redis 連線健康"""
|
||
s = DbRedisStats()
|
||
try:
|
||
from src.db.base import get_db_context
|
||
from sqlalchemy import text as sa_text
|
||
async with get_db_context() as db:
|
||
await db.execute(sa_text("SELECT 1"))
|
||
s.db_ok = True
|
||
s.db_status = "✅ 正常"
|
||
except Exception as e:
|
||
s.db_status = f"❌ {str(e)[:40]}"
|
||
|
||
try:
|
||
from src.core.redis_client import get_redis
|
||
redis = get_redis()
|
||
info = await redis.info("memory")
|
||
used_mb = int(info.get("used_memory", 0)) // (1024 * 1024)
|
||
all_keys = await redis.dbsize()
|
||
s.redis_ok = True
|
||
s.redis_key_count = all_keys
|
||
s.redis_status = f"✅ 正常 {used_mb}MB / {all_keys} keys"
|
||
except Exception as e:
|
||
s.redis_status = f"❌ {str(e)[:40]}"
|
||
return s
|
||
|
||
async def _get_pod_status(self) -> list[PodInfo]:
|
||
"""查 awoooi-prod namespace 的所有 Pod 狀態
|
||
2026-05-03 Claude Opus 4.7 + 統帥 ogt：P0 #3 加抓 STARTTIME 才能做 K8s state machine 判斷
|
||
"""
|
||
pods: list[PodInfo] = []
|
||
try:
|
||
import subprocess
|
||
r = subprocess.run(
|
||
["kubectl", "-n", "awoooi-prod", "get", "pods",
|
||
"--no-headers", "-o",
|
||
"custom-columns=NAME:.metadata.name,READY:.status.containerStatuses[0].ready,"
|
||
"STATUS:.status.phase,RESTARTS:.status.containerStatuses[0].restartCount,"
|
||
"STARTTIME:.status.startTime"],
|
||
capture_output=True, text=True, timeout=8,
|
||
)
|
||
for line in r.stdout.strip().splitlines():
|
||
parts = line.split()
|
||
if len(parts) >= 3:
|
||
name = parts[0]
|
||
ready = parts[1].lower() == "true"
|
||
status = parts[2]
|
||
restarts = int(parts[3]) if len(parts) >= 4 and parts[3].isdigit() else 0
|
||
start_time = parts[4] if len(parts) >= 5 and parts[4] != "<none>" else None
|
||
pods.append(PodInfo(
|
||
name=name, ready=ready, status=status,
|
||
restarts=restarts, start_time=start_time,
|
||
))
|
||
except Exception as e:
|
||
logger.debug("heartbeat_pod_status_failed", error=str(e))
|
||
return pods
|
||
|
||
async def _get_scanner_stats(self) -> ScannerStats:
|
||
"""查各 scanner 最後執行時間（Redis daily lock key TTL 反推）"""
|
||
stats = ScannerStats()
|
||
scanner_names = [
|
||
"capacity_forecaster", "hermes_rule_quality",
|
||
"compliance_scanner", "coverage_evaluator", "daily_report",
|
||
]
|
||
try:
|
||
from src.core.redis_client import get_redis
|
||
from src.utils.timezone import now_taipei as _now
|
||
redis = get_redis()
|
||
today = _now().date().isoformat()
|
||
for name in scanner_names:
|
||
key = f"aiops:daily_lock:{name}:{today}"
|
||
ttl = await redis.ttl(key)
|
||
if ttl > 0:
|
||
# TTL=25h 時剛跑完；剩餘 TTL 推算跑完時間
|
||
ran_at_sec = 25 * 3600 - ttl
|
||
h, m = divmod(ran_at_sec // 60, 60)
|
||
stats.last_runs[name] = f"今日 {h:02d}:{m:02d}"
|
||
else:
|
||
stats.last_runs[name] = None # 今日尚未執行
|
||
except Exception as e:
|
||
logger.debug("heartbeat_scanner_stats_failed", error=str(e))
|
||
return stats
|
||
|
||
async def _probe_telegram_bot(self) -> TelegramBotStats:
|
||
"""探測 Telegram Bot polling 狀態"""
|
||
s = TelegramBotStats()
|
||
try:
|
||
from src.core.redis_client import get_redis
|
||
redis = get_redis()
|
||
# polling leader lock
|
||
leader = await redis.get("telegram:polling:leader")
|
||
if leader:
|
||
s.polling_ok = True
|
||
s.status = f"✅ Polling 活躍 (leader: {leader.decode()[:20] if isinstance(leader, bytes) else str(leader)[:20]})"
|
||
else:
|
||
# 嘗試查最近 callback 時間（tg_msg: key 存在即有活動）
|
||
keys = await redis.keys("tg_msg:*")
|
||
if keys:
|
||
s.polling_ok = True
|
||
s.status = f"✅ 有活動 ({len(keys)} msg keys)"
|
||
else:
|
||
s.status = "⚠️ 無 polling leader key（可能重啟中）"
|
||
except Exception as e:
|
||
s.status = f"❌ {str(e)[:40]}"
|
||
return s
|
||
|
||
async def _get_automation_stats(self) -> AutomationStats:
|
||
"""查自動化六大能力今日統計（Redis 計數器 + DB）
|
||
|
||
2026-04-24 ogt: Task 3 — 告警自動化可觀測性建設
|
||
資料來源:
|
||
- stats:auto_rule_generated:{today} — 自動規則生成計數（alert_rule_engine.py 寫入）
|
||
- stats:auto_approve_rejected:{reason}:{today} — 拒絕自動審核原因分布
|
||
- knowledge_entries 今日新增（DB）
|
||
- approval_records drift_adopted 今日（DB）
|
||
"""
|
||
stats = AutomationStats()
|
||
try:
|
||
from src.core.redis_client import get_redis
|
||
redis = get_redis()
|
||
today = now_taipei().strftime("%Y%m%d")
|
||
|
||
# 自動規則生成今日計數
|
||
raw = await redis.get(f"stats:auto_rule_generated:{today}")
|
||
stats.auto_rule_generated_today = int(raw) if raw else 0
|
||
|
||
# 自動審核拒絕原因分布
|
||
reject_keys = await redis.keys(f"stats:auto_approve_rejected:*:{today}")
|
||
for key in reject_keys:
|
||
key_str = key.decode() if isinstance(key, bytes) else key
|
||
# key 格式：stats:auto_approve_rejected:{reason}:{today}
|
||
parts = key_str.split(":")
|
||
reason_part = parts[2] if len(parts) >= 4 else key_str
|
||
count_raw = await redis.get(key_str)
|
||
stats.reject_counts[reason_part] = int(count_raw) if count_raw else 0
|
||
except Exception as e:
|
||
logger.debug("heartbeat_automation_redis_failed", error=str(e))
|
||
|
||
try:
|
||
from src.db.base import get_db_context
|
||
from sqlalchemy import text as sa_text
|
||
async with get_db_context() as db:
|
||
# 今日新增 KM（timestamptz 直接比較，不需 AT TIME ZONE）
|
||
km_today = await db.scalar(sa_text(
|
||
"SELECT COUNT(*) FROM knowledge_entries "
|
||
"WHERE created_at >= NOW() - interval '24 hours'"
|
||
))
|
||
stats.km_created_today = int(km_today or 0)
|
||
|
||
# 今日 Config Drift 自動採納（查 drift_reports，adopt() 在此表更新 status）
|
||
drift_today = await db.scalar(sa_text(
|
||
"SELECT COUNT(*) FROM drift_reports "
|
||
"WHERE status = 'adopted' "
|
||
"AND resolved_at >= NOW() - interval '24 hours'"
|
||
))
|
||
stats.drift_adopted_today = int(drift_today or 0)
|
||
except Exception as e:
|
||
logger.debug("heartbeat_automation_db_failed", error=str(e))
|
||
|
||
return stats
|
||
|
||
# =========================================================================
|
||
# Warnings 彙整
|
||
# =========================================================================
|
||
|
||
def _build_warnings(self, report: HeartbeatReport) -> list[str]:
|
||
warnings: list[str] = []
|
||
|
||
# Ollama 模型未載入
|
||
for model, loaded in report.ollama_models.items():
|
||
if not loaded:
|
||
warnings.append(f"{model} 未載入，相關功能失效")
|
||
|
||
# AI 服務異常
|
||
for name, probe in report.ai_services.items():
|
||
if not probe.ok and not probe.status.startswith("⚠️"):
|
||
warnings.append(f"{name} 服務異常: {probe.status}")
|
||
|
||
# MCP 設定問題
|
||
for name, probe in report.mcp_providers.items():
|
||
if not probe.ok:
|
||
warnings.append(f"MCP {name}: {probe.status}")
|
||
|
||
# ArgoCD 非 Synced+Healthy
|
||
argocd = report.infra.get("argocd_sync")
|
||
if argocd and not argocd.ok:
|
||
warnings.append(f"ArgoCD: {argocd.status}")
|
||
|
||
# 飛輪警告
|
||
if report.flywheel.playbook_count == 0:
|
||
warnings.append("Playbook 數量為 0，飛輪學習無法啟動")
|
||
if report.flywheel.km_total > 0:
|
||
vec_rate = report.flywheel.km_vectorized / report.flywheel.km_total
|
||
if vec_rate < 0.8:
|
||
warnings.append(
|
||
f"KM 向量化率偏低: {report.flywheel.km_vectorized}/{report.flywheel.km_total}"
|
||
f" ({int(vec_rate*100)}%)"
|
||
)
|
||
|
||
# 24h 沉默（無學習活動）— 2h 太短，正常無事故期間必然誤報
|
||
if report.flywheel.last_learning_at:
|
||
silence_hours = (now_taipei() - report.flywheel.last_learning_at.replace(
|
||
tzinfo=report.timestamp.tzinfo if report.timestamp.tzinfo else None
|
||
)).total_seconds() / 3600
|
||
if silence_hours > 24:
|
||
warnings.append(f"系統沉默 {silence_hours:.1f}h（無學習活動）")
|
||
|
||
# DB / Redis 異常
|
||
if not report.db_redis.db_ok:
|
||
warnings.append(f"PostgreSQL: {report.db_redis.db_status}")
|
||
if not report.db_redis.redis_ok:
|
||
warnings.append(f"Redis: {report.db_redis.redis_status}")
|
||
|
||
# Pending 積壓告警
|
||
if report.alert_pipeline.pending_approval > 10:
|
||
warnings.append(f"PENDING 積壓 {report.alert_pipeline.pending_approval} 筆，需人工處理")
|
||
|
||
# Pod 異常 — 2026-05-03 Claude Opus 4.7 + 統帥 ogt：P0 #3 完整 K8s pod state machine
|
||
# K8s pod phases (https://kubernetes.io/docs/concepts/workloads/pods/pod-lifecycle/):
|
||
# Pending — 已建立但容器還沒起（短暫 OK，>5min 異常 = image pull / scheduling 卡）
|
||
# Running — 至少 1 容器跑中（ready=False 短暫 OK，>2min 異常 = probe fail）
|
||
# Succeeded — 全容器成功結束（CronJob/Job 正常，不算未就緒）
|
||
# Failed — 全容器結束，至少 1 fail（必告警）
|
||
# Unknown — 狀態無法取得（必告警）
|
||
from datetime import datetime, timezone
|
||
_now = datetime.now(timezone.utc)
|
||
_PENDING_THRESHOLD_MIN = 5
|
||
_NOT_READY_THRESHOLD_MIN = 2
|
||
|
||
def _age_minutes(start_time: Optional[str]) -> Optional[float]:
|
||
"""ISO 8601 startTime → 距今分鐘。None 或解析失敗返 None。"""
|
||
if not start_time:
|
||
return None
|
||
try:
|
||
# K8s startTime 格式：2026-05-03T12:34:56Z
|
||
dt = datetime.fromisoformat(start_time.replace("Z", "+00:00"))
|
||
return (_now - dt).total_seconds() / 60.0
|
||
except (ValueError, TypeError):
|
||
return None
|
||
|
||
for pod in report.pods:
|
||
phase = pod.status
|
||
age_min = _age_minutes(pod.start_time)
|
||
|
||
# restart 次數高無論 phase 都告警（CrashLoop 中或跑完都值得追）
|
||
# 放最前面，避免後面 continue 跳過
|
||
if pod.restarts >= 3:
|
||
warnings.append(f"Pod {pod.name} 重啟 {pod.restarts} 次")
|
||
|
||
if phase in ("Succeeded", "Completed"):
|
||
# CronJob/Job 成功跑完，ready=False 是設計，phase 部分不算未就緒
|
||
continue
|
||
elif phase == "Failed":
|
||
# 真正失敗 — 一定告警
|
||
warnings.append(f"Pod {pod.name} Failed")
|
||
elif phase == "Unknown":
|
||
warnings.append(f"Pod {pod.name} 狀態 Unknown")
|
||
elif phase == "Pending":
|
||
# 短暫 Pending OK；持續 >5min 表示 image pull / scheduling 卡住
|
||
if age_min is None or age_min >= _PENDING_THRESHOLD_MIN:
|
||
age_str = f"{age_min:.0f}m" if age_min else "未知"
|
||
warnings.append(f"Pod {pod.name} 持續 Pending {age_str}（image pull / scheduling 卡住）")
|
||
elif phase == "Running" and not pod.ready:
|
||
# Running 但 not ready：短暫 OK（剛起）；>2min 表示 probe fail / 啟動慢
|
||
if age_min is None or age_min >= _NOT_READY_THRESHOLD_MIN:
|
||
age_str = f"{age_min:.0f}m" if age_min else "未知"
|
||
warnings.append(f"Pod {pod.name} NotReady {age_str}（readiness probe fail / 啟動異常）")
|
||
# Running + ready=True 是健康狀態，跳過
|
||
|
||
return warnings
|
||
|
||
|
||
def report_to_telegram_html(report: HeartbeatReport) -> str:
|
||
"""
|
||
將 HeartbeatReport 轉換為 Telegram HTML 格式
|
||
|
||
ADR-075 TYPE-1 格式 (2026-04-12 ogt):
|
||
💚 INFO | AWOOOI 系統報告 + ├─/└─ 樹狀結構
|
||
"""
|
||
ts = report.timestamp.strftime("%Y-%m-%d %H:%M (台北)")
|
||
overall_ok = not report.warnings
|
||
|
||
header_icon = "💚" if overall_ok else "⚠️"
|
||
header_label = "全系統正常" if overall_ok else f"需關注 {len(report.warnings)} 項"
|
||
|
||
lines = [
|
||
f"{header_icon} <b>INFO | AWOOOI 系統報告</b>",
|
||
f"⏰ {ts}",
|
||
"──────────────────────",
|
||
"",
|
||
]
|
||
|
||
# --- AI 服務 ---
|
||
ollama = report.ai_services.get("ollama", ProbeResult(False, "❌"))
|
||
ollama_lat = f" {ollama.latency_ms:.0f}ms" if ollama.latency_ms else ""
|
||
models_ok = [m.split(":")[0] for m, ok in report.ollama_models.items() if ok]
|
||
models_str = " / ".join(models_ok) if models_ok else "無模型"
|
||
nem = report.ai_services.get("nemotron", ProbeResult(False, "❌"))
|
||
gem = report.ai_services.get("gemini", ProbeResult(False, "❌"))
|
||
cla = report.ai_services.get("claude", ProbeResult(False, "❌"))
|
||
|
||
lines.append("🤖 <b>AI 服務</b>")
|
||
lines.append(f"├─ Ollama: {ollama.status}{ollama_lat} <code>{html.escape(models_str)}</code>")
|
||
lines.append(f"├─ Nemotron NIM: {nem.status}" + (f" {nem.latency_ms:.0f}ms" if nem.latency_ms else ""))
|
||
lines.append(f"├─ Gemini API: {gem.status}" + (f" {gem.latency_ms:.0f}ms" if gem.latency_ms else ""))
|
||
lines.append(f"└─ Claude API: {cla.status}" + (f" {cla.latency_ms:.0f}ms" if cla.latency_ms else ""))
|
||
lines.append("")
|
||
|
||
# --- MCP Provider ---
|
||
k8s = report.mcp_providers.get("k8s", ProbeResult(False, "❌"))
|
||
ssh = report.mcp_providers.get("ssh", ProbeResult(False, "❌"))
|
||
argocd_mcp = report.mcp_providers.get("argocd", ProbeResult(False, "❌"))
|
||
sentry_mcp = report.mcp_providers.get("sentry", ProbeResult(False, "❌"))
|
||
|
||
lines.append("🔌 <b>MCP Provider</b>")
|
||
lines.append(f"├─ K8s: {k8s.status} SSH: {ssh.status}")
|
||
lines.append(f"└─ ArgoCD: {argocd_mcp.status} Sentry: {sentry_mcp.status}")
|
||
lines.append("")
|
||
|
||
# --- 飛輪狀態 ---
|
||
fw = report.flywheel
|
||
if fw.attempt_24h > 0:
|
||
rate = int(fw.success_24h / fw.attempt_24h * 100)
|
||
repair_str = f"{fw.success_24h}/{fw.attempt_24h} ({rate}%)"
|
||
else:
|
||
repair_str = "0 次"
|
||
km_str = ""
|
||
if fw.km_total > 0:
|
||
vec_rate = int(fw.km_vectorized / fw.km_total * 100)
|
||
km_icon = "✅" if vec_rate >= 90 else "⚠️"
|
||
km_str = f"KM: {km_icon} {fw.km_vectorized}/{fw.km_total} ({vec_rate}%)"
|
||
learn_str = f" 學習: {fw.last_learning_at.strftime('%H:%M')}" if fw.last_learning_at else ""
|
||
|
||
lines.append("🔄 <b>飛輪狀態（24h）</b>")
|
||
lines.append(f"├─ Playbooks: {fw.playbook_count} 修復: {repair_str}")
|
||
lines.append(f"└─ {km_str}{learn_str}" if km_str or learn_str else "└─ KM 統計不可用")
|
||
lines.append("")
|
||
|
||
# --- 基礎設施 ---
|
||
argocd = report.infra.get("argocd_sync", ProbeResult(False, "❌"))
|
||
velero = report.infra.get("velero", ProbeResult(False, "❌"))
|
||
|
||
lines.append("🚀 <b>基礎設施</b>")
|
||
lines.append(f"├─ ArgoCD: {argocd.status}")
|
||
lines.append(f"└─ Velero: {velero.status}")
|
||
|
||
# --- 告警流水線 ---
|
||
ap = report.alert_pipeline
|
||
lines.append("")
|
||
lines.append("📊 <b>告警流水線（24h）</b>")
|
||
lines.append(f"├─ 總計: {ap.total_24h} PENDING: {ap.pending_approval}")
|
||
if ap.execution_success_24h > 0 and ap.execution_failed_24h == 0:
|
||
exec_icon = "✅"
|
||
elif ap.execution_failed_24h > 0:
|
||
exec_icon = "⚠️"
|
||
else:
|
||
exec_icon = "—"
|
||
lines.append(f"└─ 執行: {exec_icon} 成功 {ap.execution_success_24h} 失敗 {ap.execution_failed_24h}")
|
||
|
||
# --- DB & Redis ---
|
||
dr = report.db_redis
|
||
lines.append("")
|
||
lines.append("🗄️ <b>資料庫 & Redis</b>")
|
||
lines.append(f"├─ PostgreSQL: {dr.db_status}")
|
||
lines.append(f"└─ Redis: {dr.redis_status} Keys: {dr.redis_key_count}")
|
||
|
||
# --- K8s Pods ---
|
||
if report.pods:
|
||
lines.append("")
|
||
lines.append("☸️ <b>Kubernetes Pods</b>")
|
||
for i, pod in enumerate(report.pods):
|
||
prefix = "└─" if i == len(report.pods) - 1 else "├─"
|
||
ready_icon = "✅" if pod.ready else "❌"
|
||
restart_str = f" (重啟×{pod.restarts})" if pod.restarts > 0 else ""
|
||
lines.append(f"{prefix} {ready_icon} {html.escape(pod.name[:35])}{restart_str}")
|
||
|
||
# --- Scanner 狀態 ---
|
||
if report.scanners.last_runs:
|
||
lines.append("")
|
||
lines.append("⏱️ <b>Scanner 狀態（今日）</b>")
|
||
scanner_items = list(report.scanners.last_runs.items())
|
||
for i, (name, ran_at) in enumerate(scanner_items):
|
||
prefix = "└─" if i == len(scanner_items) - 1 else "├─"
|
||
icon = "✅" if ran_at else "⏸️"
|
||
ran_str = ran_at or "尚未執行"
|
||
lines.append(f"{prefix} {icon} {html.escape(name)}: {ran_str}")
|
||
|
||
# --- Telegram Bot ---
|
||
tg = report.telegram_bot
|
||
lines.append("")
|
||
lines.append("🤖 <b>Telegram Bot</b>")
|
||
lines.append(f"└─ {tg.status}")
|
||
|
||
# --- 自動化統計（2026-04-24）---
|
||
auto = report.automation
|
||
lines.append("")
|
||
lines.append("⚙️ <b>自動化統計（今日）</b>")
|
||
lines.append(f"├─ 新規則自動生成: {auto.auto_rule_generated_today} 條")
|
||
lines.append(f"├─ KM 新增: {auto.km_created_today} 條")
|
||
lines.append(f"├─ Drift 自動採納: {auto.drift_adopted_today} 筆")
|
||
if auto.reject_counts:
|
||
reject_total = sum(auto.reject_counts.values())
|
||
top_reason = max(auto.reject_counts, key=auto.reject_counts.get) # type: ignore[arg-type]
|
||
lines.append(
|
||
f"└─ 人工審核攔截: {reject_total} 次 主因: <code>{html.escape(top_reason)}</code>"
|
||
)
|
||
else:
|
||
lines.append("└─ 人工審核攔截: 0 次")
|
||
|
||
# --- Warnings / 總結 ---
|
||
lines.append("")
|
||
if report.warnings:
|
||
lines.append(f"⚠️ <b>需關注（{len(report.warnings)} 項）</b>")
|
||
for w in report.warnings[:-1]:
|
||
lines.append(f"├─ {html.escape(w)}")
|
||
lines.append(f"└─ {html.escape(report.warnings[-1])}")
|
||
else:
|
||
lines.append(f"✅ <b>{header_label}</b>")
|
||
|
||
return "\n".join(lines)
|