---
title: "Troubleshoot Adaptive Profiles boosts | Grafana Cloud documentation"
description: "Troubleshooting guide for Adaptive Profiles boosts, covering common issues with manual and automatic boosts not starting, version detection requirements, and boosts that complete but capture too little data."
---

> For a curated documentation index, see [llms.txt](/llms.txt). For the complete documentation index, see [llms-full.txt](/llms-full.txt).

# Troubleshoot Adaptive Profiles boosts

Boosts are short periods of elevated sampling that capture more detailed profiling data for a service. Boosts can be started manually from the UI or triggered automatically when Adaptive Profiles detects a new service version.

This guide helps diagnose problems when boosts don’t start or don’t capture enough data.

## Issue 1: Boosts do not start (manual or automatic)

Boosts cannot run if Adaptive Profiles cannot identify the service or ingest profiling data correctly.

### Required conditions (for all boosts)

No boost of any kind can run unless:

- Profiling data has been flowing for several minutes
- Each profile includes a valid `service_name`
- Adaptive Profiles is enabled for your stack

### How to check

- Open the **Adaptive Profiles** UI
- Verify your service appears in the **Service** dropdown

If the service does not appear, boosts cannot be triggered.

### What to do

- Verify profiling ingestion is working
- Confirm `service_name` is included in profiles
- Ensure Adaptive Profiles is enabled

## Issue 2: Automatic boosts do not trigger

Automatic boosts depend on detecting a new service version.

### Version detection uses

Adaptive Profiles detects new versions using:

- The `BuildId` field in profile data, or
- The `service_git_ref` label

### Why automatic boosts may not run

Automatic boosts do not trigger if:

- `BuildId` does not change across deployments
- `service_git_ref` does not change across deployments
- Deployment metadata is missing

### What to do

- Ensure your CI/CD process updates **at least one** of these fields on every deployment

Manual boosts can still be triggered from the UI even if version metadata does not change. Automatic boosts require metadata to be present *and* updated over time.

## Issue 3: Boost started but captured too little data

A boost can start successfully but still fail to produce enough detailed data to be useful for performance analysis.

### Common causes

- The service generates too few samples during the boost window.
- The boost duration is too short, which is common for low-traffic services.
- Temporary backend ingestion delays.

### What to do

- Increase the boost duration, for example to 10 to 30 minutes.
- Start boosts during periods of higher service traffic.

A boost needs to capture a complete, sufficiently populated high-resolution window to be useful when you investigate the service in Drilldown.