KEDA Integration
Full observability for event-driven autoscaling. Monitor KEDA scaler health, trigger metric values, scaling event frequency, and HPA correlation in real time.
How It Works
Enable KEDA Prometheus Metrics
KEDA exposes metrics via its operator and metrics-apiserver. TigerOps deploys ServiceMonitor resources targeting both components, collecting scaler health, reconciliation latency, and trigger metric values.
Deploy TigerOps via Helm
Install the TigerOps Helm chart with the KEDA integration enabled. The chart creates RBAC to read ScaledObject, ScaledJob, and TriggerAuthentication CRDs, enriching raw metrics with scaling target context.
Map Scalers to Workloads
TigerOps automatically joins KEDA ScaledObject metrics with the metrics from the scaling target (Deployment, StatefulSet, etc.) and the upstream trigger source (Kafka topic lag, queue depth, HTTP request rate).
Configure Scaling & Scaler Alerts
Set alerts for scaler errors, scale-to-zero pauses, and slow scale-up events where trigger metric growth outpaces pod readiness. TigerOps correlates scaling lag with downstream latency impact.
What You Get Out of the Box
Scaler Health & Error Tracking
Monitor active/inactive scaler state per ScaledObject, scaler error counts, and the last successful metric fetch time. Alert immediately when a scaler loses connectivity to its trigger source (Kafka, SQS, Redis, etc.).
Scaling Event Timeline
Record every scale-up and scale-down event with the trigger metric value that caused it, the current replica count, and time to reach desired replicas. Build per-ScaledObject scaling frequency histograms.
Trigger Metric Value Monitoring
Track the raw trigger metric values (queue depth, consumer lag, RPS, CPU%) that KEDA uses to make scaling decisions. Correlate trigger metric spikes with scaling response times to identify slow autoscaler reactions.
HPA-to-KEDA Trigger Correlation
KEDA generates HPA objects for each ScaledObject. TigerOps links KEDA trigger metrics with the HPA current/desired replica counts and conditions, giving a unified view of the autoscaling pipeline from trigger to pod.
Scale-to-Zero & Cold Start Tracking
Monitor scale-from-zero events and measure cold start duration — time from first trigger activation to first pod ready. Identify workloads with unacceptable cold start latency for scale-to-zero optimization.
AI Scaling Pattern Optimization
TigerOps AI analyzes your KEDA scaling history to identify suboptimal cooldown periods, polling intervals, or activation thresholds. Recommendations surface directly in the TigerOps console with expected latency improvements.
TigerOps Helm Values for KEDA Monitoring
Configure ServiceMonitors and CRD enrichment for comprehensive KEDA observability.
# TigerOps Helm values for KEDA integration
# helm repo add tigerops https://charts.atatus.net
# helm install tigerops tigerops/tigerops -f values.yaml
global:
apiKey: "${TIGEROPS_API_KEY}"
remoteWriteEndpoint: https://ingest.atatus.net/api/v1/write
keda:
enabled: true
namespace: keda
# Scrape KEDA operator metrics
operator:
metricsPort: 8080
scrapeInterval: 15s
# Scrape KEDA metrics apiserver
metricsApiserver:
metricsPort: 9022
scrapeInterval: 15s
# Enrich with ScaledObject and ScaledJob CRD data
crdEnrichment:
enabled: true
includeScaledJobs: true
# HPA correlation — join KEDA trigger metrics with HPA replicas
hpaCorrelation:
enabled: true
# Scale-to-zero cold start tracking
scaleToZero:
coldStartTracking: true
alertColdStartSeconds: 30
alerts:
scalerErrorCount: 3
scalerInactiveMinutes: 60
scalingResponseSeconds: 45
triggerMetricStalenessSeconds: 60Common Questions
Which KEDA trigger types does TigerOps support for metric monitoring?
TigerOps collects the generic keda_scaler_metrics_value metric that KEDA exposes for all trigger types. Additionally, for common triggers (Kafka, AWS SQS, Azure Service Bus, Redis, Prometheus), TigerOps correlates the trigger source metrics directly to provide full end-to-end visibility.
How does TigerOps handle ScaledJobs differently from ScaledObjects?
TigerOps monitors ScaledJobs with job-specific metrics: max replica count (max job parallelism), active job count, completed job rate, and job execution duration histograms. ScaledJobs are tracked separately from ScaledObjects in TigerOps dashboards.
Can TigerOps alert when a KEDA scaler is paused or inactive for too long?
Yes. TigerOps tracks the keda_scaled_object_paused metric and fires an alert when a ScaledObject is paused longer than a configured threshold. It also alerts when a scaler has been inactive (metric value at zero) for an anomalously long period based on its historical pattern.
Does TigerOps integrate with TriggerAuthentication secrets for scaler monitoring?
TigerOps does not read TriggerAuthentication secret values. It monitors TriggerAuthentication object status conditions (ready/not ready) and correlates authentication failures with scaler error spikes, helping diagnose credential expiry without exposing secrets.
How does TigerOps measure the lag between a trigger metric breach and pod readiness?
TigerOps calculates scaling response time by tracking three timestamps: when the trigger metric crossed the activation threshold, when the HPA scaled the deployment, and when the new pods became Ready. This end-to-end latency is reported per ScaledObject in the scaling latency dashboard.
Stop Guessing Why Your Event-Driven Scaling Is Slow
Trigger metric values, scaler health, cold start latency, and HPA correlation — the complete KEDA observability picture in one dashboard.