핵심 질문
- metric은 application에서 Prometheus까지 어떻게 노출·수집·저장되는가?
- PromQL은 어떤 time-series question에 답하고, Grafana는 그 결과를 어떻게 탐색하게 하는가?
- Alertmanager는 언제 alert delivery와 routing을 담당하는가?
읽기 경로
- Prometheus
- Prometheus Exporter
- PromQL
- Grafana
- Loki
- Alertmanager
- 장애 자동 대응의 안전장치
Build·delivery 경로
- Docker Platform
- Docker Build Cache Design
- GitHub Actions Workflow
메트릭·로그·경보의 흐름
flowchart LR
T[Application / host / target]
E[Exporter or instrumented app]
P[Prometheus]
G[Grafana]
R[Alert rule]
M[Alertmanager]
C[Receiver]
T -->|메트릭 노출| E
P -->|메트릭 scrape 요청| E
G -->|PromQL 조회 요청| P
G -->|LogQL 조회 요청| L[Loki]
T -->|로그 읽기 대상| A[Promtail]
A -->|로그 push| L
R -->|평가 규칙 설정| P
P -->|규칙 평가로 생성한 경보 전달| M
M -->|그룹화와 라우팅 뒤 알림 전달| C
C -->|조치 요청 검증과 실행 정책 적용| GR[Guarded Alert Remediation]현재 범위
- 현재 source는 Prometheus monitoring stack과 Docker·GitHub Actions의 build/delivery boundary를 함께 다룬다.
- Prometheus, Alertmanager, Grafana의 공식 문서를 함께 확인해 configuration, alert lifecycle, datasource/variable 모델을 보강했다.
- Docker는 client-daemon, image/container, volume/network lifecycle까지, GitHub Actions는 event-triggered workflow·job·step·runner model까지 다룬다.
- Dockerfile detail, build-cache rule, Compose syntax, reusable workflow·secret forwarding recipe는 추가 official source가 생길 때 확장한다.
blackbox_exporter는 target의 /metrics를 전달하는 exporter가 아니라 target을 probe하고 probe 결과를 노출하는 예외적 흐름이다.