/grafana-billing

Grafana Billing Metrics Reference

This document defines the key metrics used for observability billing and how they're calculated.

Prometheus Metrics

Active Time Series

Metric: prometheus_tsdb_head_series

The number of unique time series currently in Prometheus's head block (recent data in memory).

• Billing Impact: Primary billing dimension for Grafana Cloud Metrics
• Calculation: 95th percentile over billing period (forgives ~36 hours of spikes per month)
• Optimization: Reduce label cardinality, drop unused metrics

Data Points Per Minute (DPM)

Metric: rate(prometheus_tsdb_head_samples_appended_total[5m]) * 60

The rate at which new data points are being ingested.

• Billing Impact: Secondary billing dimension
• Calculation: DPM = samples/second × 60
• Optimization: Increase scrape interval, reduce metric count

Head Chunks

Metric: prometheus_tsdb_head_chunks

Number of chunks in the head block. Each time series has multiple chunks.

• Billing Impact: Memory usage indicator
• Normal Ratio: ~2-3 chunks per active series

TSDB Storage

Metric: prometheus_tsdb_storage_blocks_bytes

Total on-disk storage used by all TSDB blocks.

• Billing Impact: Storage costs
• Factors: Retention period, series count, sample rate

Cardinality Analysis

Endpoint: /api/v1/status/tsdb

Returns breakdown of series count by:

• seriesCountByMetricName - Which metrics have most series
• labelValueCountByLabelName - Which labels have most unique values
• memoryInBytesByLabelName - Memory cost per label

Loki Metrics

Bytes Received

Metric: loki_distributor_bytes_received_total

Cumulative bytes ingested by Loki distributors.

• Billing Impact: Primary billing dimension (GB ingested)
• Labels: tenant for multi-tenant deployments

Ingestion Rate

Calculation: rate(loki_distributor_bytes_received_total[5m])

Current ingestion rate in bytes/second.

• Conversion: GB/day = bytes/sec × 86400 / (1024³)
• Billing: Grafana Cloud charges per GB ingested

Active Streams

Metric: loki_ingester_memory_streams

Number of active log streams (unique label combinations).

• Billing Impact: Affects query performance, not direct billing
• Optimization: Reduce unique label values

Memory Chunks

Metric: loki_ingester_memory_chunks

Chunks held in memory by ingesters.

• Billing Impact: Memory usage, not direct billing
• Optimization: Tune chunk_idle_period, chunk_target_size

Rejected Bytes

Metric: loki_distributor_bytes_received_total{reason=~".+"}

Bytes rejected due to rate limiting or validation errors.

• Billing Impact: Not billed, but indicates problems
• Common Reasons: Rate limiting, line too long, stream limit

Grafana Cloud Billing Model

Metrics (Prometheus)

Logs (Loki)

Cost Optimization Strategies

Prometheus

1. Reduce cardinality: Remove high-cardinality labels (UUIDs, timestamps)
2. Drop unused metrics: Use relabeling to filter at scrape time
3. Increase scrape interval: 30s → 60s halves DPM
4. Recording rules: Pre-aggregate expensive queries

Loki

1. Drop debug logs: Filter verbose logs before ingestion
2. Compress logs: Use structured logging with templates
3. Reduce label cardinality: Static labels only
4. Use Adaptive Logs: Automatically identify droppable patterns

References

• Grafana Cloud Pricing
• Analyze Metrics Usage
• Reduce Logs Costs
• Prometheus TSDB