What causes REQUEST_THROTTLED?

Too many concurrent API calls to the Databricks Jobs, Clusters, or SQL API from automation or tooling. An orchestration platform (Airflow, ADF, custom scripts) polling job status at too high a frequency. Multiple automated pipelines or monitoring tools hammering the Databricks REST API simultaneously. A Databricks SQL warehouse being queried by too many concurrent Power BI connections. An API client not implementing exponential backoff, causing retry storms

How do I fix REQUEST_THROTTLED?

Implement exponential backoff with jitter in all Databricks API clients. Reduce the polling frequency for job status checks — use webhooks or callbacks instead of polling where possible. Use the Databricks SDK which includes built-in retry and rate limit handling. Review concurrent API usage across all tools — consolidate status polling into a single monitoring layer. For SQL warehouse throttling, consider upgrading the warehouse size or enabling serverless SQL

Medium severityapi

Power BI Refresh Error:
REQUEST_THROTTLED

What does this error mean?

A request to the Databricks API or an external service was rate-limited. The request was rejected due to exceeding the allowed request rate — no data was lost, but the operation must be retried.

Common causes

1Too many concurrent API calls to the Databricks Jobs, Clusters, or SQL API from automation or tooling
2An orchestration platform (Airflow, ADF, custom scripts) polling job status at too high a frequency
3Multiple automated pipelines or monitoring tools hammering the Databricks REST API simultaneously
4A Databricks SQL warehouse being queried by too many concurrent Power BI connections
5An API client not implementing exponential backoff, causing retry storms

How to fix it

1Implement exponential backoff with jitter in all Databricks API clients
2Reduce the polling frequency for job status checks — use webhooks or callbacks instead of polling where possible
3Use the Databricks SDK which includes built-in retry and rate limit handling
4Review concurrent API usage across all tools — consolidate status polling into a single monitoring layer
5For SQL warehouse throttling, consider upgrading the warehouse size or enabling serverless SQL

Frequently asked questions

What are Databricks API rate limits?

Databricks applies rate limits per workspace and per API endpoint. The exact limits are published in the Databricks documentation and vary by API. The Jobs API, Clusters API, and SQL API each have separate limits.

Does REQUEST_THROTTLED cause data loss?

No — throttled requests are rejected before execution. Retry the request after a delay. Implement exponential backoff to avoid compounding the throttling with immediate retries.