KEDA

KEDA Integration in Dash0

KEDA (Kubernetes Event-Driven Autoscaling) is a lightweight component that brings event-driven autoscaling to Kubernetes by managing Horizontal Pod Autoscalers behind the scenes. Unlike traditional HPA scaling based on CPU and memory, KEDA can scale workloads based on events and external signals like message queue depth, database job tables, or custom application metrics.

This integration enables comprehensive observability of KEDA's control plane through OpenTelemetry metrics while demonstrating how to use Dash0's Prometheus-compatible API to drive intelligent autoscaling decisions.

For more information, visit the KEDA Project Page.

Use Cases

Transform your Kubernetes autoscaling strategy with event-driven patterns:

• Observable Autoscaling: Monitor KEDA's internal metrics to understand scaling decisions and troubleshoot autoscaler behavior
• Custom Metrics Scaling: Scale workloads based on application-specific metrics from Dash0, moving beyond CPU and memory
• Scale-to-Zero: Efficiently scale workloads down to zero replicas when no work is waiting, optimizing resource usage
• Event-Driven Workloads: Respond to external signals like queue depth, database job counts, or scheduled events
• Multi-Trigger Scaling: Combine multiple scaling triggers for complex autoscaling logic

Setup

This guide shows how to set up KEDA with OpenTelemetry observability and Dash0 integration. For a complete working example, see the Dash0 KEDA demo repository.

1. Install KEDA with OpenTelemetry

Install KEDA with OpenTelemetry metrics enabled:

Create a values file for KEDA:

Install KEDA:

2. Configure Authentication

Create a Kubernetes secret with your Dash0 API token:

Create a TriggerAuthentication resource:

3. Example: HTTP Request Rate Scaling

Create a ScaledObject that scales based on HTTP request rate from Dash0:

Configure the OpenTelemetry Collector

KEDA sends telemetry data to an OpenTelemetry Collector, which forwards it to Dash0.

For production environments, consider using one of these deployment methods:

• Dash0 operator for Kubernetes (recommended)
• OpenTelemetry Operator for Kubernetes

• Helm Chart for The OpenTelemetry Collector

The key requirement is that your collector accepts OTLP/HTTP on port 4318.

Next Steps

• Explore the KEDA scalers documentation for more trigger types
• Set up alerts based on KEDA control plane metrics
• Create custom dashboards combining KEDA metrics with application metrics
• Experiment with multiple triggers on a single ScaledObject for complex scaling logic