feat(infra): Tempo + Prometheus + Grafana observability backplane in compose (refs #122)

Three built images on the cg network with config baked in: Tempo (OTLP ingest,
ports 4317/4318), Prometheus, and Grafana with both datasources auto-provisioned.
Config lives in infra/observability/. The verify-observability step runs early in
CI verify-stack; the three containers are added to the failure log-dump list.
Tempo's Grafana plugin has no datasource /health method, so the check proves
Tempo reachability through Grafana's datasource proxy instead.

refs #122
This commit is contained in:
not
2026-07-23 12:46:54 +02:00
parent f12e9ce746
commit f29a8598fd
9 changed files with 120 additions and 9 deletions
+39
View File
@@ -524,6 +524,45 @@ services:
condition: service_started
networks: [cg]
# ── Observability backplane (S-16a, ADR-0023) ──────────────────────────────
# Grafana-native stack: Tempo ingests OTLP traces (the .NET services export
# straight to it — no collector hop, S-16b), Prometheus scrapes service
# /metrics (S-16c), and Grafana reads both with datasources auto-provisioned.
# Config is baked into small built images (COPY) rather than streamed into
# external config volumes like the upstream CG modules — these aren't verbatim
# peer images, so a built image is the simpler path that still reaches sibling
# containers on the CI runner. Not in WAIT_SVCS: run-observability-check.sh
# polls Grafana itself, so no in-image healthcheck tool is needed.
tempo:
build:
context: ./observability/tempo
image: register-referentie/tempo:dev
command: ["-config.file=/etc/tempo.yaml"]
networks: [cg]
prometheus:
build:
context: ./observability/prometheus
image: register-referentie/prometheus:dev
ports:
- "9090:9090"
networks: [cg]
grafana:
build:
context: ./observability/grafana
image: register-referentie/grafana:dev
environment:
GF_SECURITY_ADMIN_USER: admin
GF_SECURITY_ADMIN_PASSWORD: admin
GF_AUTH_ANONYMOUS_ENABLED: "true"
ports:
- "3000:3000"
depends_on:
- tempo
- prometheus
networks: [cg]
volumes:
oz-db:
nrc-db: