specification: API Commons Rate Limits
specificationVersion: '0.1'
schema: https://raw.githubusercontent.com/api-evangelist/interface-research/main/schema/api-commons.yml#/$defs/RateLimits
provider: Dynatrace
providerId: dynatrace
created: '2026-05-04'
modified: '2026-08-29'
generated: '2026-08-29'
method: searched
source: https://docs.dynatrace.com/docs/dynatrace-api/basics/access-limit
reconciled: true
tags:
  - Rate Limiting
  - Observability
  - APM
description: >-
  Dynatrace throttles the API by CAPACITY, not by a published request-per-second quota. Every
  environment has a limited thread pool with a queue; you hit the limit when both the pool and its
  queue are full, or when a request times out waiting in the queue after 10 seconds. Exhaustion
  returns HTTP 429. What Dynatrace DOES publish precisely is payload ceilings per API, and a
  charging model in which reads are free on fair use while defining and pushing custom metrics is
  billed per metric per month.
notes: >-
  Corrected 2026-08-29. The prior version of this artifact was written by the 2026-05-04 bulk sweep
  with method:generated and asserted X-RateLimit-Limit / X-RateLimit-Remaining / X-RateLimit-Reset
  and Retry-After response headers, a 503 status, and a "sandbox vs production" limit tier. A
  direct read of the Dynatrace access-limit page and a scan of every captured OpenAPI found NONE
  of those — no rate-limit header of any kind is documented or declared, 503 appears nowhere, and
  Dynatrace publishes no sandbox. Those claims are removed rather than carried forward.
sources:
  - https://docs.dynatrace.com/docs/dynatrace-api/basics/access-limit
  - https://docs.dynatrace.com/docs/dynatrace-api/basics/dynatrace-api-response-codes

headers: {}
headers_note: >-
  No X-RateLimit-*, RateLimit-* or Retry-After response header is documented on the access-limit
  page, and none is declared on any operation in the captured OpenAPI set. An agent receiving a 429
  from Dynatrace gets a status code and nothing else — no budget, no reset time, no retry hint — so
  it must back off exponentially with jitter on its own judgement.

responseCodes:
  throttled: 429
  throttledMeaning: >-
    Too many requests. The environment's request thread pool AND its queue are full, or the request
    timed out in the queue (10-second timeout).

limits:
  - name: Environment request throttling (thread pool + queue)
    scope: environment
    metric: concurrent_requests
    limit: 'capacity-based — not a published numeric quota'
    timeFrame: instantaneous
    queue_timeout_seconds: 10
    status_on_exhaustion: 429
    notes: >-
      Deliberately designed so many cheap requests pass while expensive ones are the ones that
      queue. Cost per request, not request count, is what exhausts the limit.
  - name: Default payload size
    scope: request
    metric: bytes
    limit: 1048576
    limit_human: 1 MB
    timeFrame: request
  - name: Log Monitoring API payload
    scope: request
    metric: bytes
    limit: 10485760
    limit_human: 10 MB
    timeFrame: request
  - name: Ingestion API payload (upper tier)
    scope: request
    metric: bytes
    limit: 8388608
    limit_human: 8 MB
    timeFrame: request
    notes: Ingestion endpoints cap at 8 MB, 4 MB or 2 MB depending on the endpoint.
  - name: Extensions API upload
    scope: request
    metric: bytes
    limit: 52428800
    limit_human: 50 MB ZIP
    timeFrame: request
  - name: Plugins API upload
    scope: request
    metric: bytes
    limit: 52428800
    limit_human: 50 MB ZIP
    timeFrame: request
  - name: Mobile Symbolication API upload
    scope: request
    metric: bytes
    limit: 104857600
    limit_human: 100 MiB compressed / 500 MiB uncompressed
    timeFrame: request
  - name: Custom metric charging
    scope: environment
    metric: custom_metrics_per_month
    limit: 'billed per metric per month — governed by DPS or the classic DDU model, not by HTTP rate'
    timeFrame: month
    notes: >-
      Read access to the Dynatrace API is free of charge on a fair-use model. Defining and pushing
      NEW custom metrics is charged.

limit_count: 8

policies:
  - name: Back off blindly on 429
    description: >-
      There is no Retry-After header to honour. Retry with exponential backoff and jitter, and
      widen the interval rather than the batch — the throttle is capacity-based, so a smaller,
      cheaper request is more likely to pass than a delayed large one.
  - name: Split expensive queries
    description: >-
      Because the queue admits many cheap requests and few expensive ones, narrowing a selector or
      shortening a query window is a more effective remedy than reducing request frequency.
  - name: Retries are not safe on ingest
    description: >-
      Dynatrace publishes no idempotency key. A retry after a 429 on ingestLogs, ingestEvent or
      ingestCustomMetrics duplicates the datapoint. See conventions/dynatrace-conventions.yml.
  - name: Subscription-driven quotas are separate from HTTP throttling
    description: >-
      Ingest volume is bounded by the Dynatrace Platform Subscription or the classic DDU/host
      entitlement. That is a licence limit, not an HTTP rate limit, and it does not surface as 429.

maintainers:
  - FN: Kin Lane
    email: kin@apievangelist.com