prometheus（Prometheus 监控）

Prometheus Querying

Setup

Prometheus and Alertmanager are behind OAuth2 Proxy on the internal gateway. DNS URLs (https://prometheus.internal.tomnowak.work) redirect to an OAuth login page and cannot be used for API queries via curl.

Access Methods

Option 1: kubectl exec (quick, no setup)

# Query Prometheus API directly inside the pod
kubectl --context <cluster> exec -n monitoring prometheus-kube-prometheus-stack-0 -c prometheus -- \
  wget -qO- 'http://localhost:9090/api/v1/query?query=up' | jq '.data.result'

# Query Alertmanager API
kubectl --context <cluster> exec -n monitoring alertmanager-kube-prometheus-stack-0 -c alertmanager -- \
  wget -qO- 'http://localhost:9093/api/v2/alerts' | jq .

Option 2: Port-forward (for scripts and repeated queries)

kubectl --context <cluster> port-forward -n monitoring svc/prometheus-operated 9090:9090 &
export PROMETHEUS_URL=http://localhost:9090

The promql.sh script defaults to http://localhost:9090 and works with port-forward out of the box.

Quick Queries

Use the bundled script at .claude/skills/prometheus/scripts/promql.sh:

# Start port-forward first (script defaults to http://localhost:9090)
kubectl --context <cluster> port-forward -n monitoring svc/prometheus-operated 9090:9090 &

# Instant query
./scripts/promql.sh query 'up'

# Range query (last hour, 15s resolution)
./scripts/promql.sh range 'node_cpu_seconds_total' --start 1h --step 15s

# Firing alerts only
./scripts/promql.sh alerts --firing

# All alert rules
./scripts/promql.sh rules

# Find metrics by label
./scripts/promql.sh series '{job="node-exporter"}'

# Health check
./scripts/promql.sh health

Common Queries

# CPU usage %
./scripts/promql.sh query 'avg(1 - rate(node_cpu_seconds_total{mode="idle"}[5m])) * 100'

# Memory usage %
./scripts/promql.sh query '(1 - sum(node_memory_MemAvailable_bytes) / sum(node_memory_MemTotal_bytes)) * 100'

# Problem pods
./scripts/promql.sh query 'kube_pod_status_phase{phase!~"Running|Succeeded"} == 1'

# Container restarts (last hour)
./scripts/promql.sh query 'increase(kube_pod_container_status_restarts_total[1h]) > 0'

# Firing alerts
./scripts/promql.sh alerts --firing

# Full alert details
./scripts/promql.sh alerts --verbose

Creating Monitoring Resources

This skill covers querying Prometheus. For authoring new monitoring resources (PrometheusRules, ServiceMonitors, PodMonitors, AlertmanagerConfig, recording rules), see the monitoring-authoring skill.

Prometheus Details