mgmt/infrastructure/monitor/configs/prometheus/prometheus.yml

62 lines
1.7 KiB
YAML

# Prometheus 配置 - 监控Nomad集群
global:
scrape_interval: 15s
evaluation_interval: 15s
external_labels:
monitor: 'nomad-cluster'
# Alertmanager configuration
alerting:
alertmanagers:
- static_configs:
- targets: ['localhost:9093']
# Load rules once and periodically evaluate them according to the global 'evaluation_interval'.
rule_files:
# - "first_rules.yml"
# - "second_rules.yml"
# A scrape configuration containing exactly one endpoint to scrape:
scrape_configs:
# Prometheus自身监控
- job_name: 'prometheus'
scrape_interval: 5s
scrape_timeout: 5s
static_configs:
- targets: ['localhost:9090']
# Node Exporter - 客户端节点
- job_name: 'node-clients'
static_configs:
- targets:
- 'ch4.tailnet-68f9.ts.net:9100'
- 'ash3c.tailnet-68f9.ts.net:9100'
- 'warden.tailnet-68f9.ts.net:9100'
- 'hcp1.tailnet-68f9.ts.net:9100'
- 'browser.tailnet-68f9.ts.net:9100'
# Node Exporter - 服务端节点
- job_name: 'node-servers'
static_configs:
- targets:
- 'ash2e.tailnet-68f9.ts.net:9100'
- 'ch2.tailnet-68f9.ts.net:9100'
- 'ch3.tailnet-68f9.ts.net:9100'
- 'onecloud1.tailnet-68f9.ts.net:9100'
# Nomad集群监控
- job_name: 'nomad'
static_configs:
- targets:
- 'ash1.tailnet-68f9.ts.net:4646'
- 'ash2.tailnet-68f9.ts.net:4646'
- 'onecloud1.tailnet-68f9.ts.net:4646'
# Consul集群监控
- job_name: 'consul'
static_configs:
- targets:
- 'ash1.tailnet-68f9.ts.net:8500'
- 'ash2.tailnet-68f9.ts.net:8500'
- 'onecloud1.tailnet-68f9.ts.net:8500'