mirror of
git://f0xx.org/ac/ac-deploy
synced 2026-07-29 06:17:57 +03:00
d045653a7b
artc0 VM is off; remove node_exporter target and relabel rules from Prometheus. Point Grafana auth-proxy whitelist and alert-notifier shortener API at cast01–03 cluster instead of 10.7.16.128. Co-authored-by: Cursor <cursoragent@cursor.com>
118 lines
4.3 KiB
YAML
118 lines
4.3 KiB
YAML
# Prometheus main config — cast04 (10.7.16.239)
|
|
# Track A monitoring SPEC: ac-docs/specs/20260626_cluster_monitoring.md
|
|
# Updated: 2026-06-28 — mysqld targets moved to cast01/02/03 cluster
|
|
|
|
global:
|
|
scrape_interval: 15s
|
|
evaluation_interval: 15s
|
|
external_labels:
|
|
cluster: 'ac-cluster0'
|
|
monitor: 'cast04'
|
|
|
|
rule_files:
|
|
- "/etc/prometheus/rules/*.yml"
|
|
|
|
alerting:
|
|
alertmanagers:
|
|
- static_configs:
|
|
- targets: ['localhost:9093']
|
|
|
|
scrape_configs:
|
|
|
|
# ── node_exporter on all cluster nodes ──────────────────────────────────────
|
|
- job_name: 'node'
|
|
static_configs:
|
|
- targets:
|
|
- '10.7.16.236:9100' # cast01
|
|
- '10.7.16.237:9100' # cast02
|
|
- '10.7.16.238:9100' # cast03
|
|
- '10.7.16.239:9100' # cast04 (self)
|
|
labels:
|
|
env: 'production'
|
|
relabel_configs:
|
|
- source_labels: [__address__]
|
|
regex: '(10\.7\.16\.236):.*'
|
|
target_label: instance
|
|
replacement: 'cast01'
|
|
- source_labels: [__address__]
|
|
regex: '(10\.7\.16\.237):.*'
|
|
target_label: instance
|
|
replacement: 'cast02'
|
|
- source_labels: [__address__]
|
|
regex: '(10\.7\.16\.238):.*'
|
|
target_label: instance
|
|
replacement: 'cast03'
|
|
- source_labels: [__address__]
|
|
regex: '(10\.7\.16\.239):.*'
|
|
target_label: instance
|
|
replacement: 'cast04'
|
|
|
|
# ── blackbox_exporter — internal HTTP probes (no TLS issues) ────────────────
|
|
- job_name: 'blackbox_http'
|
|
metrics_path: /probe
|
|
params:
|
|
module: [http_2xx]
|
|
static_configs:
|
|
- targets:
|
|
- 'http://10.7.16.236/app/androidcast_project/' # BE main app (cast01)
|
|
- 'http://10.7.16.236/app/androidcast_project/issues/' # crash reporter
|
|
- 'http://10.7.16.239:3000/api/health' # Grafana health
|
|
labels:
|
|
env: 'production'
|
|
relabel_configs:
|
|
- source_labels: [__address__]
|
|
target_label: __param_target
|
|
- source_labels: [__param_target]
|
|
target_label: instance
|
|
- target_label: __address__
|
|
replacement: 'localhost:9115'
|
|
|
|
# ── blackbox_exporter — external HTTPS probes (self-signed cert on FE) ──────
|
|
# Uses insecure_skip_verify until a valid LE cert is deployed on apps.f0xx.org
|
|
# TODO: switch back to http_2xx once cert is fixed (FE monstro: certbot --nginx)
|
|
- job_name: 'blackbox_https_external'
|
|
metrics_path: /probe
|
|
params:
|
|
module: [http_2xx_insecure]
|
|
static_configs:
|
|
- targets:
|
|
- 'https://apps.f0xx.org/app/androidcast_project/' # external FE (self-signed cert)
|
|
# notify.f0xx.org: add here once FE nginx proxies notify.f0xx.org → cluster
|
|
labels:
|
|
env: 'production'
|
|
relabel_configs:
|
|
- source_labels: [__address__]
|
|
target_label: __param_target
|
|
- source_labels: [__param_target]
|
|
target_label: instance
|
|
- target_label: __address__
|
|
replacement: 'localhost:9115'
|
|
|
|
# ── MariaDB exporter — cast01 (primary) ────────────────────────────────────
|
|
- job_name: 'mysqld'
|
|
static_configs:
|
|
- targets: ['10.7.16.236:9104']
|
|
labels:
|
|
instance: 'cast01-mariadb'
|
|
|
|
# ── PostgreSQL exporter (cast01) ─────────────────────────────────────────────
|
|
- job_name: 'postgres'
|
|
static_configs:
|
|
- targets: ['10.7.16.236:9187']
|
|
labels:
|
|
instance: 'cast01-postgres'
|
|
|
|
# ── prometheus self-scrape ───────────────────────────────────────────────────
|
|
- job_name: 'prometheus'
|
|
static_configs:
|
|
- targets: ['localhost:9090']
|
|
labels:
|
|
instance: 'cast04-prometheus'
|
|
|
|
# ── alertmanager self-scrape ─────────────────────────────────────────────────
|
|
- job_name: 'alertmanager'
|
|
static_configs:
|
|
- targets: ['localhost:9093']
|
|
labels:
|
|
instance: 'cast04-alertmanager'
|