Mimir (Long-Term Metrics)

Grafana Mimir provides long-term metric storage with a 90-day retention period, extending well beyond Prometheus's 15-day local retention. Prometheus writes all metrics to Mimir via remote write, and Grafana queries Mimir as an additional Prometheus-compatible data source.

Architecture

Component Layout

Mimir runs in distributed mode with single replicas per component, all pinned to the VPS node:

Component	Replicas	Storage	Resources
Distributor	1	—	50m–500m CPU, 128–512Mi
Ingester	1	10Gi local-path	100m–500m CPU, 256Mi–1Gi
Compactor	1	10Gi local-path	50m–500m CPU, 128–512Mi
Store Gateway	1	10Gi local-path	50m–500m CPU, 128–512Mi
Querier	1	—	50m–500m CPU, 128–512Mi
Query Frontend	1	—	50m–200m CPU, 64–256Mi

Disabled components (not needed for single-node): query_scheduler, ruler, alertmanager, overrides_exporter.

Configuration

Storage Backend

All Mimir components use MinIO as the S3-compatible storage backend:

mimir:
  structuredConfig:
    common:
      storage:
        backend: s3
        s3:
          endpoint: prometheus-minio.monitoring.svc.cluster.local:9000
          insecure: true
    blocks_storage:
      s3:
        bucket_name: mimir-blocks
      tsdb:
        dir: /data/tsdb
    ruler_storage:
      s3:
        bucket_name: mimir-ruler
    alertmanager_storage:
      s3:
        bucket_name: mimir-alertmanager

Retention

limits:
  compactor_blocks_retention_period: 2160h  # 90 days

Ingestion Limits

Rate limits are disabled for single-tenant operation:

limits:
  max_global_series_per_user: 0     # unlimited
  ingestion_rate: 0                  # unlimited
  ingestion_burst_size: 0            # unlimited

Replication

All replication factors are set to 1 for single-node deployment:

ingester:
  ring:
    replication_factor: 1
store_gateway:
  sharding_ring:
    replication_factor: 1

Data Flow

1. Prometheus writes all scraped metrics to Mimir's gateway endpoint via remoteWrite
2. Distributor receives and forwards samples to the ingester
3. Ingester batches samples into TSDB blocks on local disk
4. Compactor merges and deduplicates blocks, uploads to MinIO, enforces 90-day retention
5. Store Gateway serves historical blocks from MinIO for queries
6. Querier fans out queries to both the ingester (recent data) and store gateway (historical data)
7. Grafana queries Mimir through the query frontend at http://prometheus-mimir-gateway.monitoring.svc.cluster.local/prometheus