429 Too Many Requests from the Datadog API: back off with jitter, then reduce calls

TL;DR: Datadog answers 429 Too Many Requests when you exceed API rate limits. Back off, honor Retry-After, and stop hammering the endpoint that throttled you. Stop and wait. Add exponential backoff with jitter to the client.

429 Too Many Requests: back off, dont retry harder

What you see: API calls start returning 429 with an error body saying too many requests. Your Terraform apply, your custom exporter, or your script loop suddenly fails after working fine.

Why it happens

Datadog rate-limits API endpoints per organization. The usual triggers:

• A script polling an endpoint in a tight loop (metrics queries every second from ten hosts).
• Terraform provider runs against hundreds of monitors in parallel.
• A retry loop without backoff: the first 429 causes immediate retries, which extends the throttle.

Fix

1. Stop and wait. Honor the Retry-After header if present. A fixed 60 second pause beats hammering.
2. Add exponential backoff with jitter to the client. First retry after a few seconds, doubling, plus randomness so ten hosts do not retry in lockstep.
3. Reduce call volume. Cache responses, batch requests, widen poll intervals. If you query the same dashboard data every 10s from CI, you probably need a monitor, not a script.
4. Split across keys/endpoints only if the limit is per-key; most are per-org, so this rarely helps. Check which.

The agent is not this

Agent intake throttling is separate from API 429s. If the agent is being throttled, you will see it in datadog-agent status as dropped payloads, and the fix is usually fewer custom metrics or a support ticket, not client backoff. Do not confuse the two.

Trap

Treating 429 as an auth failure and rotating keys. It is not auth. It is pace. Rotating keys mid-throttle just adds a second problem.

When to use this

• This covers exactly what the title says: 429 Too Many Requests from the Datadog API.
• You are setting this up for the first time, or auditing an existing setup.
• You want the key gotchas in one place before you start.

When not to use this

• You are doing a different workflow with 429; these steps are specific to the title above.
• You need the full reference docs; this is the short path, not the manual.

Compatibility

• Not pinned to a specific version; follows current 429 behavior.

The API reference documents 429 Too many requests as a standard response code (e.g. the Monitors API lists 200/400/401/429 with 429 meaning too many requests). Rate limiting applies per endpoint and per org. Verified against live docs 2026-09-26.