Loki SLIs

This page documents Service Level Indicators (SLIs) for monitoring the health of Loki. To set up Service Level Objectives (SLOs), see Set up SLOs with Sloth.

These metrics are recommended as Service Level Indicators for Loki. They cover the write path (log ingestion), read path (query serving), and supporting subsystems.

Log ingestion (write path)

The distributor receives log streams from Promtail or other clients, and the ingester persists them to object storage. These metrics measure the health of the write path.

Metric	Type	Description
loki_distributor_lines_received_total	Counter	Total log lines received by the distributor
loki_distributor_bytes_received_total	Counter	Total bytes received by the distributor
loki_distributor_ingester_appends_total	Counter	Total successful append calls to ingesters
loki_discarded_samples_total	Counter	Samples discarded by reason (rate_limited, out_of_order, ingestion_rate, etc.)
loki_discarded_bytes_total	Counter	Bytes discarded by reason

An availability SLI for the write path is the ratio of successful ingester appends:

sum(rate(loki_distributor_ingester_appends_total[5m]))

A discard rate SLI shows the proportion of data being rejected:

sum by (reason) (rate(loki_discarded_samples_total[5m]))

Ingester

The ingester holds recent log streams in memory, builds chunks, and flushes them to the backend store. These metrics capture ingester health and resource pressure.

Metric	Type	Description
loki_ingester_memory_streams	Gauge	Number of active streams held in ingester memory
loki_ingester_memory_chunks	Gauge	Number of chunks in ingester memory
loki_ingester_chunks_flushed_total	Counter	Total chunks flushed to backend storage
loki_ingester_chunks_flush_failures_total	Counter	Chunk flush failures; should remain at 0
loki_ingester_flush_queue_length	Gauge	Number of chunks queued for flushing; high values indicate backend pressure

A flush failure SLI:

sum(rate(loki_ingester_chunks_flush_failures_total[5m]))
/
sum(rate(loki_ingester_chunks_flushed_total[5m]))

Write-ahead log (WAL)

The WAL provides durability for unflushed data in the ingester.

Metric	Type	Description
loki_ingester_wal_disk_full_failures_total	Counter	Failures due to WAL disk being full
loki_ingester_wal_replay_active	Gauge	Whether WAL replay is currently active (1) or not (0)

Query performance (read path)

Loki serves log queries through the query frontend, queriers, and query scheduler. These metrics measure read-path health.

Metric	Type	Description
loki_request_duration_seconds	Histogram	HTTP/gRPC request latency by route, method, and status code
loki_inflight_requests	Gauge	Number of inflight requests across all routes
loki_query_frontend_queries_in_progress	Gauge	Number of queries currently being processed
loki_query_frontend_retries	Histogram	Number of retries per query
loki_query_scheduler_queue_duration_seconds	Histogram	Time queries spend in the scheduler queue before being picked up
loki_querier_worker_inflight_queries	Gauge	Inflight queries per worker

A latency SLI for the read API (P99):

histogram_quantile(0.99, sum by (le) (
  rate(loki_request_duration_seconds_bucket{method="gRPC",route!~".*Pusher.*"}[5m])
))

TSDB index

Loki uses TSDB-based index storage. These metrics capture index build and maintenance health.

Metric	Type	Description
loki_tsdb_build_index_attempts_total	Counter	TSDB index build attempts by status (success, failure)
loki_tsdb_build_index_last_successful_timestamp_seconds	Gauge	Timestamp of the last successful index build
loki_tsdb_head_rotation_attempts_total	Counter	TSDB head rotation attempts by status
loki_tsdb_wal_truncation_attempts_total	Counter	WAL truncation attempts by status

An index build success SLI:

sum by (component) (rate(loki_tsdb_build_index_attempts_total{status="success"}[5m]))
/
sum by (component) (rate(loki_tsdb_build_index_attempts_total[5m]))

Cache

Loki uses caching to accelerate queries and reduce storage load. Cache efficiency directly impacts read-path latency.

Metric	Type	Description
loki_cache_hits	Counter	Total cache hits
loki_cache_fetched_keys	Counter	Total cache keys fetched
loki_cache_corrupt_chunks_total	Counter	Corrupt chunks detected in cache

A cache hit ratio SLI:

sum(rate(loki_cache_hits[5m]))
/
sum(rate(loki_cache_fetched_keys[5m]))

Ruler (alerting)

Loki’s ruler evaluates recording and alerting rules against LogQL expressions.

Metric	Type	Description
loki_ruler_config_last_reload_successful	Gauge	Whether the last ruler config reload succeeded (1 = success, 0 = failure)
loki_prometheus_notifications_errors_total	Counter	Alert notification errors
loki_prometheus_notifications_dropped_total	Counter	Alert notifications dropped

Ring / cluster membership

Loki uses a distributed hash ring for component discovery and ownership.

Metric	Type	Description
loki_ring_members	Gauge	Ring members by name (ingester, distributor, compactor, scheduler) and state (ACTIVE, JOINING, LEAVING, PENDING, Unhealthy)

A ring health SLI (unhealthy members should be 0):

sum by (name) (loki_ring_members{state="Unhealthy"})

Compactor

The compactor merges TSDB index files and applies retention.

Metric	Type	Description
loki_compactor_apply_retention_operation_duration_seconds	Gauge	Duration of retention application
loki_compactor_apply_retention_last_successful_run_timestamp_seconds	Gauge	Timestamp of the last successful retention run
loki_boltdb_shipper_compact_tables_operation_total	Counter	Table compaction operations by status