Improve performance by mitigating rate limited request errors

This page describes how to use Storage Intelligence advisor to help improve performance by mitigating rate limited request errors.

To provide service stability, Cloud Storage limits request rates on buckets and objects. Exceeding these limits can cause 429 Too Many Requests errors, leading to application latency and timeouts.

When rate-limited request errors increase, Storage Intelligence advisor displays a Spike in 429 errors indicating rate limited requests finding. A finding is an anomaly detected from your metrics that needs your attention. For a full list of findings, see Findings reference catalog.

Thresholds that trigger a finding

Storage Intelligence creates a baseline for your project, folder, or organization based on historical trends. Storage Intelligence advisor generates a finding when 429 errors increase above the baseline.

Findings reflect the previous day's activity. For example, a spike occurring on Tuesday appears in Storage Intelligence advisor on Wednesday.

Respond to a finding

  1. View the finding details.

  2. Use the following table to select the recommended strategy for your scenario:

    Scenario Recommended action Details
    Cause of 429 spike is unclear Use Gemini Cloud Assist to identify the issue If the cause and scale of the issue are unclear, use Gemini Cloud Assist to analyze logs and identify specific application or user trends causing the overload. For setup details, see Analyze stored data with Gemini Cloud Assist.
    Client application issues Optimize client behavior Review your application code to verify that it implements exponential backoff. Consider distributing workloads across multiple buckets or prefixes to prevent exceeding per-bucket limits.
    Need granular request logs Investigate access patterns with Storage Insights datasets

    Export metadata to BigQuery for a detailed breakdown of object-level access patterns. This information can help you identify specific patterns or objects that are contributing to contention.

What's next