generated: '2026-08-27'
method: searched
source: >-
  https://docs.lakekeeper.io/docs/latest/configuration/, https://docs.lakekeeper.io/docs/latest/monitoring/,
  https://docs.lakekeeper.io/docs/latest/admission/, https://docs.lakekeeper.io/llms-full.txt
limit_count: 0
note: >-
  Lakekeeper publishes NO rate limits and returns NO rate-limit response headers, and that is a property of the
  product rather than a documentation gap: it is self-hosted server software, so throughput is bounded by the
  operator's own Postgres, object store and pod count, not by a vendor quota. A full-text search of the published
  documentation corpus (llms-full.txt, 580KB) finds no X-RateLimit-*, no RateLimit-*, and no documented 429
  response from Lakekeeper itself. Every 429 mentioned in the docs is one Lakekeeper RECEIVES from an upstream
  (Microsoft Graph, cloud STS/SAS endpoints) and retries with Retry-After — the client side, not the server side.
limits: []
response_headers:
  published: []
  note: No X-RateLimit-*, RateLimit-* or Retry-After header is documented on a Lakekeeper response.
exhaustion_status: null
request_guards:
- name: LAKEKEEPER__MAX_REQUEST_BODY_SIZE
  default: 2 MB
  description: Rejects oversized request bodies. Introduced in v0.12.0.
  source: https://docs.lakekeeper.io/docs/latest/configuration/
- name: LAKEKEEPER__MAX_REQUEST_TIME
  default: 30s
  description: Rejects slow requests. Introduced in v0.12.0.
  source: https://docs.lakekeeper.io/docs/latest/configuration/
- name: LAKEKEEPER__PAGINATION_SIZE_MAX
  default: not stated
  description: >-
    Caps page size on list operations. Setting it to 4294967295 (u32::MAX) restores the Iceberg spec's
    return-everything behaviour when pageToken is unset.
  source: https://docs.lakekeeper.io/docs/latest/configuration/
- name: LAKEKEEPER__MAINTENANCE_MODE
  default: 'off'
  description: 'Set to `read-only` to refuse writes server-wide. Captured at startup, not dynamic.'
adjacent_controls:
- name: Admission gates
  description: >-
    A pluggable post-authentication hook that can deny a request before it reaches a handler — the closest thing
    Lakekeeper has to per-tenant throttling. A gate returns a terminal 403, or 503 with Retry-After when it fails
    closed. Gates run on every authenticated request and are expected to cache aggressively.
  source: https://docs.lakekeeper.io/docs/latest/admission/
- name: Endpoint statistics
  description: >-
    Per-project, per-endpoint API call counts, readable via POST /management/v1/endpoint-statistics
    (get_endpoint_statistics) and routable to a custom EndpointStatisticsSink. This is the metering surface an
    operator would build chargeback or quotas on — Lakekeeper measures, the operator enforces.
  source: https://docs.lakekeeper.io/docs/latest/monitoring/
outbound_retry_behaviour:
  description: >-
    As a CLIENT, Lakekeeper retries transient upstream failures (429 honouring Retry-After, transient 5xx,
    connection/timeout errors) a few times with exponential backoff before falling back to the last good cached
    result. Applies to identity-provider role lookups and cloud STS/SAS credential minting.
  source: https://docs.lakekeeper.io/llms-full.txt