generated: '2026-08-27'
method: searched
source: https://grafana.com/docs/loki/latest/operations/request-validation-rate-limits/
name: Grafana Loki rate limits
description: >-
  Grafana Loki is self-hosted software, so every rate limit is operator-configured per tenant rather
  than fixed by a vendor. Loki publishes both the limit semantics and the shipped default value for
  each one in its configuration reference, which is what an agent needs to predict a 429 before it
  happens. Limits are enforced per tenant (identified by the X-Scope-OrgID header) by the
  distributor and ingester components.
enforcement:
  scope: per-tenant
  tenant_header: X-Scope-OrgID
  components: [distributor, ingester]
  status_on_exhaustion: 429
  status_on_validation_failure: 400
  status_on_oversize_body: 413
response_headers:
  published: false
  note: >-
    Loki does not document any X-RateLimit-* or RateLimit-* response headers and does not document a
    Retry-After header on 429. The runtime signal Loki does publish is Prometheus metrics: the
    counters loki_discarded_samples_total and loki_discarded_bytes_total carry a `reason` label whose
    values are the identifiers in errors/grafana-loki-problem-types.yml, scraped from GET /metrics.
    An agent cannot read its remaining budget from a response; it must observe 429 + reason and back
    off. This is a real gap in the runtime signal.
  observed_headers: []
  retry_after: false
rate_limits:
  - name: ingestion_rate_mb
    scope: per-tenant
    window: per-second
    limit: 4
    unit: MB/s
    burst: 6
    burst_unit: MB
    burst_config: ingestion_burst_size_mb
    status: 429
    reason: rate_limited
    enforced_by: distributor
    retryable: true
    configurable_per_tenant: true
    note: Shipped default. Operators routinely raise this.
  - name: per_stream_rate_limit
    scope: per-stream
    window: per-second
    limit: 3
    unit: MB/s
    burst: 15
    burst_unit: MB
    burst_config: per_stream_rate_limit_burst
    status: 429
    reason: per_stream_rate_limit
    enforced_by: ingester
    retryable: true
    configurable_per_tenant: true
  - name: max_global_streams_per_user
    scope: per-tenant
    window: concurrent
    limit: 5000
    unit: active streams
    status: 429
    reason: stream_limit
    enforced_by: ingester
    retryable: true
    configurable_per_tenant: true
    note: 'An active stream is one that received a line within chunk_idle_period (default 30m).'
  - name: max_line_size
    scope: per-log-line
    window: per-request
    limit: 262144
    unit: bytes
    limit_display: 256KB
    status: 400
    reason: line_too_long
    enforced_by: distributor
    retryable: false
    configurable_per_tenant: true
  - name: max_label_names_per_series
    scope: per-stream
    window: per-request
    limit: 15
    unit: labels
    status: 400
    reason: max_label_names_per_series
    enforced_by: distributor
    retryable: false
    configurable_per_tenant: true
  - name: max_label_name_length
    scope: per-label
    window: per-request
    limit: 1024
    unit: bytes
    status: 400
    reason: label_name_too_long
    enforced_by: distributor
    retryable: false
    configurable_per_tenant: true
  - name: max_label_value_length
    scope: per-label
    window: per-request
    limit: 2048
    unit: bytes
    status: 400
    reason: label_value_too_long
    enforced_by: distributor
    retryable: false
    configurable_per_tenant: true
  - name: max_structured_metadata_size
    scope: per-log-line
    window: per-request
    limit: 65536
    unit: bytes
    limit_display: 64KB
    status: 400
    reason: structured_metadata_too_large
    enforced_by: distributor
    retryable: false
    configurable_per_tenant: true
  - name: max_structured_metadata_entries_count
    scope: per-log-line
    window: per-request
    limit: 128
    unit: entries
    status: 400
    reason: structured_metadata_too_many
    enforced_by: distributor
    retryable: false
    configurable_per_tenant: true
  - name: creation_grace_period
    scope: per-sample
    window: timestamp
    limit: 600
    unit: seconds
    limit_display: 10m
    status: 400
    reason: too_far_in_future
    enforced_by: distributor
    retryable: false
    configurable_per_tenant: true
  - name: reject_old_samples_max_age
    scope: per-sample
    window: timestamp
    limit: 604800
    unit: seconds
    limit_display: 1w
    status: 400
    reason: greater_than_max_sample_age
    enforced_by: distributor
    retryable: false
    configurable_per_tenant: true
    note: 'Only enforced while reject_old_samples is true (the shipped default).'
  - name: max_entries_limit_per_query
    scope: per-query
    window: per-request
    limit: 5000
    unit: log entries
    status: 400
    reason: null
    enforced_by: querier
    retryable: false
    configurable_per_tenant: true
  - name: max_query_series
    scope: per-query
    window: per-request
    limit: 500
    unit: series
    status: 400
    reason: null
    enforced_by: querier
    retryable: false
    configurable_per_tenant: true
  - name: max_query_length
    scope: per-query
    window: query time range
    limit: 2595600
    unit: seconds
    limit_display: 30d1h
    status: 400
    reason: null
    enforced_by: query-frontend
    retryable: false
    configurable_per_tenant: true
limit_count: 14
cloud_note: >-
  Grafana Cloud Logs (the hosted Loki service) applies its own per-plan limits on top of these; the
  Free tier is capped at 50 GB ingested per month. See plans/grafana-loki-plans-pricing.yml.
docs:
  - https://grafana.com/docs/loki/latest/operations/request-validation-rate-limits/
  - https://grafana.com/docs/loki/latest/configure/
checked: '2026-08-27'