An online retail platform running microservices on Google Kubernetes Engine (GKE) with Cloud Service Mesh triggers an alert due to a violation of its user-facing latency Service Level Objective (SLO). A DevOps engineer investigates the incident in Cloud Monitoring Metrics Explorer and correlates the following multi-dimensional golden signal metrics across tiers: Frontend External Application Load Balancer: A sharp increase in https/total_latencies at the 95th percentile, with steady request volume. Intermediate Microservices: Envoy sidecar access logs report increased upstream timeouts and 5xx status codes when querying backend dependencies. Backend Workload: CPU utilization metrics show severe saturation at 100%, accompanied by a high server error rate (5xx HTTP responses). How should the DevOps engineer interpret these correlated metrics to identify the root cause of the degradation?

The backend service experienced resource saturation (CPU exhaustion), which degraded its processing capacity, leading to backend 5xx errors and upstream response delays that cascaded to the frontend load balancer.: ### What the metric correlation indicates Correlating **golden signals** across architectural boundaries allows engineers to trace the causal relationship between underlying infrastructure constraints and user-facing SLO degradation. When analyzing multi-tier distributed applications, **saturation metrics** (such as container CPU utilization) often represent the foundational catalyst for downstream performance anomalies. ### How it meets the causal analysis requirements - **Establishing Root Cause**: The backend workload's CPU utilization reached **100% saturation**, impairing its thread execution and query-processing throughput. - **Propagating Error Rates**: As CPU exhaustion occurred, backend processes queued or timed out, generating **server error rates (5xx HTTP responses)**. - **Upstream Latency Impact**: The intermediate service proxies logged upstream timeouts while waiting on the saturated backend, which directly inflated the total round-trip processing time recorded as elevated **P95 `https/total_latencies`** on the external load balancer. - **SLO Breach**: The overall processing delay directly breached the user-facing latency SLO, despite traffic remaining constant. ### Key advantages of multi-dimensional correlation - **Pinpoints Point of Origin**: Isolates whether latency originates at the edge proxy, network layer, or deep backend computational tier. - **Eliminates Traffic Anomaly False Leads**: Verifying consistent request counts confirms the degradation was caused by internal resource starvation rather than an external distributed denial-of-service (DDoS) or traffic spike. - **Differentiates Proxy vs. Workload Issues**: Cross-referencing sidecar proxy `RESPONSE_FLAGS` and upstream status codes with workload CPU reveals whether the proxy layer or application backend is failing. ### Why this is optimal Analyzing telemetry in a structured top-down or bottom-up flow—connecting **saturation**, **error rate**, and **latency**—establishes a definitive causal chain. This accurately targets remediation efforts (such as horizontal pod autoscaling or query optimization) directly at the backend microservice.

Metrics Explorer Golden Signals Root Cause Analysis | Professional Cloud DevOps Engineer

Metrics Explorer Golden Signals Root Cause Analysis

Worked example. The correct answer is already marked and every option is explained below, so there is nothing to select here. To answer questions yourself, start the free trial.

An online retail platform running microservices on Google Kubernetes Engine (GKE) with Cloud Service Mesh triggers an alert due to a violation of its user-facing latency Service Level Objective (SLO).

A DevOps engineer investigates the incident in Cloud Monitoring Metrics Explorer and correlates the following multi-dimensional golden signal metrics across tiers:

Frontend External Application Load Balancer: A sharp increase in https/total_latencies at the 95th percentile, with steady request volume.
Intermediate Microservices: Envoy sidecar access logs report increased upstream timeouts and 5xx status codes when querying backend dependencies.
Backend Workload: CPU utilization metrics show severe saturation at 100%, accompanied by a high server error rate (5xx HTTP responses).

How should the DevOps engineer interpret these correlated metrics to identify the root cause of the degradation?

The external Application Load Balancer experienced a frontend network throughput bottleneck, which delayed packet delivery to clients and created backpressure that saturated the backend CPU.

An unmanaged sudden spike in external user traffic overloaded the Istio ingress gateway, resulting in misrouted requests and false-positive CPU saturation metrics on the backend.

A Cloud Service Mesh DNS resolution failure occurred at the sidecar proxy level, preventing healthy backend instance groups from registering with the external load balancer.

The backend service experienced resource saturation (CPU exhaustion), which degraded its processing capacity, leading to backend 5xx errors and upstream response delays that cascaded to the frontend load balancer.

Professional Cloud DevOps Engineer

Metrics Explorer Golden Signals Root Cause Analysis

Related questions