loki

Best Practices

• Use TSDB store type (not BoltDB) for new deployments — it is the current recommended index store
• Configure schema_config with appropriate period_config entries and schema version (v13 current)
• Set retention_period in limits_config and enable retention in compactor for log lifecycle management
• Define ingestion rate limits (ingestion_rate_mb, ingestion_burst_size_mb, per_stream_rate_limit) to prevent overload
• Use chunk_target_size (1.5MB recommended) for optimal storage efficiency and query performance
• Enable compactor with retention_enabled for automatic log compaction and deletion
• Set auth_enabled to true for multi-tenant clusters and configure tenant_id in Promtail clients
• Configure query limits (max_query_series, max_entries_limit_per_query) to prevent expensive queries
• Use Promtail pipeline stages (json, regex, labels, timestamp) for structured log extraction
• Set positions file in Promtail to track read positions across restarts
• Use S3 or GCS object storage for chunks in production — filesystem only for development
• Configure replication_factor > 1 in production for ingester high availability
• Use relabel_configs in Promtail to derive meaningful labels from discovery metadata
• Keep label cardinality low — high-cardinality labels (user_id, request_id) degrade performance significantly