4af6487ba6
All checks were successful
CD Pipeline / workflow-shape (push) Successful in 2s
CD Pipeline / cancel-stale-cd (push) Has been skipped
CD Pipeline / tests (push) Successful in 2m31s
CD Pipeline / build-and-deploy (push) Successful in 15m18s
AWOOOI Harbor 110 Local Repair / workflow-shape (push) Successful in 1s
AWOOOI Harbor 110 Local Repair / harbor-110-local-repair (push) Successful in 22s
CD Pipeline / post-deploy-checks (push) Successful in 5m35s
252 lines
7.2 KiB
YAML
252 lines
7.2 KiB
YAML
global:
|
||
scrape_interval: 15s
|
||
evaluation_interval: 15s
|
||
external_labels:
|
||
monitor: "wooo-aiops-monitor"
|
||
environment: "production"
|
||
|
||
alerting:
|
||
alertmanagers:
|
||
- static_configs:
|
||
- targets: ["alertmanager:9093"]
|
||
|
||
rule_files:
|
||
- "alerts.yml"
|
||
|
||
scrape_configs:
|
||
- job_name: "prometheus"
|
||
static_configs:
|
||
- targets: ["localhost:9090"]
|
||
labels:
|
||
host: "110"
|
||
|
||
# === Node Exporters (5 台主機) ===
|
||
- job_name: "node-exporter-110"
|
||
scrape_interval: 30s
|
||
scrape_timeout: 25s
|
||
static_configs:
|
||
- targets: ["192.168.0.110:9100"]
|
||
labels:
|
||
host: "110"
|
||
role: "devops"
|
||
|
||
- job_name: "node-exporter-112"
|
||
static_configs:
|
||
- targets: ["192.168.0.112:9100"]
|
||
labels:
|
||
host: "112"
|
||
role: "kali-security"
|
||
|
||
- job_name: "node-exporter-188"
|
||
scrape_interval: 30s
|
||
scrape_timeout: 25s
|
||
static_configs:
|
||
- targets: ["192.168.0.188:9100"]
|
||
labels:
|
||
host: "188"
|
||
role: "ai-web"
|
||
|
||
# === Blackbox HTTP 探測 ===
|
||
- job_name: "blackbox-http"
|
||
metrics_path: /probe
|
||
params:
|
||
module: [http_2xx]
|
||
static_configs:
|
||
- targets:
|
||
- https://aiops.wooo.work
|
||
- https://mo.wooo.work
|
||
- http://192.168.0.110:3001
|
||
- http://192.168.0.110:8080/api/v1/health
|
||
- http://192.168.0.125:32334/api/v1/health
|
||
- http://192.168.0.125:32335
|
||
- https://www.tsenyang.com
|
||
- http://stock.wooo.work
|
||
- https://bitan.wooo.work
|
||
# BEGIN AWOOOI OLLAMA CLOUD EDGE PROBES
|
||
- targets:
|
||
- http://34.143.170.20:11434/api/tags
|
||
labels:
|
||
service: ollama-cloud-edge
|
||
provider: ollama_gcp_a
|
||
probe_origin: host110
|
||
boundary: cloud_edge_availability_only
|
||
- targets:
|
||
- http://34.21.145.224:11434/api/tags
|
||
labels:
|
||
service: ollama-cloud-edge
|
||
provider: ollama_gcp_b
|
||
probe_origin: host110
|
||
boundary: cloud_edge_availability_only
|
||
# END AWOOOI OLLAMA CLOUD EDGE PROBES
|
||
relabel_configs:
|
||
- source_labels: [__address__]
|
||
target_label: __param_target
|
||
- source_labels: [__param_target]
|
||
target_label: instance
|
||
- target_label: __address__
|
||
replacement: 192.168.0.110:9115
|
||
|
||
# === Blackbox TCP 探測 ===
|
||
- job_name: "blackbox-tcp"
|
||
metrics_path: /probe
|
||
params:
|
||
module: [tcp_connect]
|
||
static_configs:
|
||
- targets:
|
||
# K3s 層
|
||
- 192.168.0.125:6443
|
||
- 192.168.0.125:32334
|
||
- 192.168.0.125:32335
|
||
# 110 服務
|
||
- 192.168.0.110:9090
|
||
- 192.168.0.110:9093
|
||
- 192.168.0.110:3001
|
||
- 192.168.0.110:9000
|
||
- 192.168.0.110:5000
|
||
- 192.168.0.110:3100
|
||
# 188 服務（Ollama 已於 2026-05-06 自 AWOOOI provider pool 退役）
|
||
- 192.168.0.188:5432
|
||
- 192.168.0.188:6380
|
||
- 192.168.0.188:8089
|
||
# K3s Worker
|
||
- 192.168.0.125:32334
|
||
- 192.168.0.125:32335
|
||
relabel_configs:
|
||
- source_labels: [__address__]
|
||
target_label: __param_target
|
||
- source_labels: [__param_target]
|
||
target_label: instance
|
||
- target_label: __address__
|
||
replacement: 192.168.0.110:9115
|
||
|
||
# === cAdvisor (Docker 容器監控) ===
|
||
- job_name: "cadvisor-110"
|
||
scrape_interval: 60s
|
||
scrape_timeout: 55s
|
||
static_configs:
|
||
- targets: ["192.168.0.110:9180"]
|
||
labels:
|
||
host: "110"
|
||
|
||
# === ClawBot ===
|
||
- job_name: "blackbox-clawbot"
|
||
metrics_path: /probe
|
||
params:
|
||
module: [http_2xx]
|
||
static_configs:
|
||
- targets:
|
||
- http://192.168.0.188:8088/health
|
||
relabel_configs:
|
||
- source_labels: [__address__]
|
||
target_label: __param_target
|
||
- source_labels: [__param_target]
|
||
target_label: instance
|
||
- target_label: __address__
|
||
replacement: 192.168.0.110:9115
|
||
|
||
- job_name: "clawbot"
|
||
scrape_interval: 30s
|
||
metrics_path: /metrics
|
||
static_configs:
|
||
- targets: ["192.168.0.188:8088"]
|
||
labels:
|
||
host: "188"
|
||
service: "clawbot"
|
||
|
||
# === Gitea native metrics (GitHub freeze replacement, 2026-07-16) ===
|
||
- job_name: 'gitea-native'
|
||
scrape_interval: 60s
|
||
metrics_path: /metrics
|
||
static_configs:
|
||
- targets: ['192.168.0.110:3001']
|
||
labels:
|
||
source: 'gitea'
|
||
|
||
# === AWOOOI API Metrics (Phase O Wave A.2, 2026-04-02) ===
|
||
- job_name: 'awoooi-api'
|
||
scrape_interval: 30s
|
||
metrics_path: /metrics
|
||
static_configs:
|
||
- targets: ['192.168.0.125:32334']
|
||
labels:
|
||
host: '125'
|
||
service: 'awoooi-api'
|
||
env: 'prod'
|
||
|
||
# BEGIN AWOOOI ALERTMANAGER RUNTIME DELIVERY TRUTH
|
||
# === Alertmanager delivery truth (2026-07-16) ===
|
||
# Scrape Alertmanager's own monotonic delivery counters. The API NodePort
|
||
# load-balances across process-local counters and must not be used to compute
|
||
# webhook error rates.
|
||
- job_name: 'alertmanager-runtime'
|
||
scrape_interval: 30s
|
||
metrics_path: /metrics
|
||
static_configs:
|
||
- targets: ['alertmanager:9093']
|
||
labels:
|
||
host: '110'
|
||
service: 'alertmanager'
|
||
env: 'prod'
|
||
# Production Alertmanager exposes notification counters by integration,
|
||
# not by receiver. The committed config has exactly one webhook receiver
|
||
# (awoooi-webhook), so stamp that independently tested source contract onto
|
||
# only the two webhook counter families at scrape time.
|
||
metric_relabel_configs:
|
||
- source_labels: [__name__, integration]
|
||
separator: ';'
|
||
regex: 'alertmanager_notification_requests(_failed)?_total;webhook'
|
||
target_label: receiver_contract
|
||
replacement: 'awoooi-webhook'
|
||
# END AWOOOI ALERTMANAGER RUNTIME DELIVERY TRUTH
|
||
|
||
# === ADR-074 M1: 飛輪健康度指標 (2026-04-12 ogt) ===
|
||
- job_name: 'awoooi-flywheel'
|
||
scrape_interval: 5m
|
||
metrics_path: /api/v1/stats/flywheel/metrics
|
||
static_configs:
|
||
- targets: ['192.168.0.125:32334']
|
||
labels:
|
||
host: '125'
|
||
service: 'awoooi-flywheel'
|
||
env: 'prod'
|
||
|
||
# === ADR-074 M2: 主機間網路連通性 (2026-04-12 ogt) ===
|
||
- job_name: 'host-connectivity'
|
||
scrape_interval: 60s
|
||
metrics_path: /probe
|
||
params:
|
||
module: [tcp_connect]
|
||
static_configs:
|
||
- targets:
|
||
- 192.168.0.110:22
|
||
- 192.168.0.188:22
|
||
- 192.168.0.120:6443
|
||
- 192.168.0.121:6443
|
||
relabel_configs:
|
||
- source_labels: [__address__]
|
||
target_label: __param_target
|
||
- source_labels: [__param_target]
|
||
target_label: instance
|
||
- target_label: __address__
|
||
replacement: 192.168.0.188:9115
|
||
|
||
# === Sprint 5.2 Plan B: PostgreSQL Exporter (2026-04-08 Claude Sonnet 4.6) ===
|
||
- job_name: 'postgres-exporter'
|
||
scrape_interval: 30s
|
||
static_configs:
|
||
- targets: ['192.168.0.188:9187']
|
||
labels:
|
||
host: '188'
|
||
service: 'postgresql'
|
||
layer: 'systemd-188'
|
||
|
||
# === Sprint 5.2 Plan B: Redis Exporter (2026-04-08 Claude Sonnet 4.6) ===
|
||
- job_name: 'redis-exporter'
|
||
scrape_interval: 30s
|
||
static_configs:
|
||
- targets: ['192.168.0.188:9121']
|
||
labels:
|
||
host: '188'
|
||
service: 'redis'
|
||
layer: 'systemd-188'
|