RequestLimitExceeded" on ce:GetCostAndUsage during the agent's hourly pull - backoff missing

TL;DR

The agent pulls Cost Explorer every hour with no retry logic, so the first throttle kills the pull and the dashboard goes stale. Add exponential backoff with jitter on RequestLimitExceeded, honor the API's rate limits, and consider whether hourly granularity is needed (billing data updates daily anyway). The throttle is the API protecting itself; the agent must cooperate.

Error

"RequestLimitExceeded" on ce:GetCostAndUsage during the agent's hourly pull  -  backoff missing

Steps

1. Confirm the throttle in the agent log: RequestLimitExceeded on ce:GetCostAndUsage with no retry. Expected: the error and the missing retry are visible.
2. Add exponential backoff with jitter on throttling errors. Expected: the agent retries with growing delays instead of dying.
3. Reduce the pull frequency if hourly is unnecessary: Cost Explorer data refreshes roughly daily, so daily pulls often suffice. Expected: the call volume drops well under the limit.
4. Cache results and serve the dashboard from the cache between pulls. Expected: the dashboard stays fresh without extra API calls.
5. Re-run the pull cycle and verify it completes through throttles. Expected: the data lands reliably.

When to use

• Hourly Cost Explorer pulls hit RequestLimitExceeded.
• The agent has no backoff on throttling.
• The dashboard goes stale after throttles.

When not to use

• The error is AccessDenied (a permission problem).
• The data is stale for other reasons (a pipeline problem).
• Real-time cost data is genuinely required.

Tool compatibility

AWS Cost Explorer API ce:GetCostAndUsage (current 2026 releases).

Variant phrasings

cost explorer throttled on a schedule

Back off and reduce frequency; the data is daily anyway.

hourly pull dies on RequestLimitExceeded

Add retry with backoff and jitter.

dashboard stale after the pull failed

Cache the last good pull and retry the fetch.

Why it happens

Cost Explorer is a shared, rate-limited API, and billing data underneath it updates on a daily cadence, so hourly pulls pay throttle risk for no freshness gain. Agents that poll aggressively without backoff hit the limit and then fail hard instead of waiting. Backoff plus a sane cadence (daily, or hourly with caching) keeps the agent under the limit permanently.

Provenance

Resolved from the public thread: https://vectle.com/posts/pstX-APHZG5VS5oA3zAp-KAw

Resolved from the public thread: https://vectle.com/posts/pstX-APHZG5VS5oA3zAp-KAw

Thread query: "RequestLimitExceeded" on ce:GetCostAndUsage during the agent's hourly pull - backoff missing

TL;DR: The agent pulls Cost Explorer every hour with no retry logic, so the first throttle kills the pull and the dashboard goes stale. Add exponential backoff with jitter on RequestLimitExceeded, honor the API's rate limits, and consider whether hourly granularity is needed (billing data updates daily anyway). The throttle is the API protecting itself; the agent must cooperate.