RetroDash // TRAFFIC

For traffic metrics you need:

• node_exporter — network metrics (interface rx/tx)
• http_requests_total — HTTP request counter
• container_network_* — if using Docker/Kubernetes
• Or custom app metrics (e.g: requests_per_second)

Rate interval

Same as the Latency dashboard: the $rate_interval variable defaults to 5m. For homelabs with sparse traffic, set it to 15m or 30m so request rate panels show smooth data instead of empty windows.

Ingress controller metric mapping

The dashboard uses generic metric names. Map them to your actual ingress controller below:

Generic Metric	Traefik	Nginx Ingress	Istio
http_requests_total	traefik_entrypoint_requests_total	nginx_ingress_controller_requests	istio_requests_total
status label	code	status	response_code
http_request_duration_seconds_*	traefik_service_request_duration_seconds_*	nginx_ingress_controller_request_duration_seconds_*	istio_request_duration_milliseconds_*

Requests/sec Stat

Current number of requests per second.

Default query (HTTP requests):

sum(rate(http_requests_total[5m]))

If you use Nginx:

rate(nginx_http_requests_total[5m])

If you use Traefik:

rate(traefik_entrypoint_requests_total[5m])

If you use API Gateway (Kong, etc):

rate(kong_requests_total[5m])

Connections Gauge

Number of active connections.

Base query:

sum(node_tcp_connection_states{state="established"})

For Kubernetes:

sum(container_network_tcp_usage_total) / 1000

Filter by specific interface:

sum(node_network_transmit_bytes_total{device="eth0"})

Request Rate — Time Series

Evolution of requests per second over time.

Multi-series query (by HTTP method):

sum by (method) (rate(http_requests_total[5m]))

By status code:

sum by (status) (rate(http_requests_total[5m]))

By service/instance:

sum by (job) (rate(http_requests_total[5m]))

Change granularity: Edit [5m] → [1m] (more responsive) or [15m] (more smoothed)

Throughput — Network In/Out

Incoming and outgoing bandwidth in bytes.

Network In (received):

sum(rate(node_network_receive_bytes_total[5m])) / 1024 / 1024

Network Out (sent):

sum(rate(node_network_transmit_bytes_total[5m])) / 1024 / 1024

Note: Dividing by 1024/1024 converts bytes to MB. For Gbps, use /1000/1000/125

Exclude virtual interfaces (docker0, veth*):

sum(rate(node_network_receive_bytes_total{device!~"docker.*|veth.*"}[5m]))

Network by Node

Table with traffic for each node.

Query with breakdown by instance:

sum by (instance) (rate(node_network_receive_bytes_total[5m]))

Change grouping:

• By service: sum by (job, instance)
• By zone: sum by (zone, instance)
• By interface: sum by (device, instance)

Filter specific nodes:

sum by (instance) (rate(node_network_receive_bytes_total{instance=~"prod.*"}[5m]))

Top Talkers

Hosts/ips with highest traffic volume.

Default query:

topk(10, sum by (remote_addr) (rate(http_requests_total[5m])))

For Docker/Kubernetes (container_network_*):

topk(10, sum by (pod_name) (rate(container_network_transmit_bytes_total[5m])))

By domain (if you have a label):

topk(10, sum by (domain) (rate(http_requests_total[5m])))

Change top N: Replace topk(10 with topk(20 or topk(5

Filter by IP range:

topk(10, sum by (remote_addr) (rate(http_requests_total{remote_addr=~"192\\.168\\..*"}[5m])))

1. Go to Dashboards → Import
2. Copy the TRAFFIC dashboard JSON
3. Verify Prometheus datasource is selected
4. Import and save
5. Check that all metrics are available in your Prometheus
6. If a metric is missing, edit the query based on your exporter

Detect traffic spikes

rate(http_requests_total[5m]) >
  avg_over_time(rate(http_requests_total[5m])[1h:5m]) * 1.5

Alert if rate exceeds 150% of the last hour average.

Request/byte ratio (efficiency)

sum(rate(http_requests_total[5m])) /
  (sum(rate(node_network_transmit_bytes_total[5m])) / 1024)

Shows how many requests per KB of data transmitted (higher = more efficient).

Bandwidth saturation prediction

predict_linear(node_network_transmit_bytes_total[1h], 3600)

Predicts the value in 1 hour, useful for detecting uncontrolled growth.