9 Essential Metrics for Effective Microservice Monitoring

To monitor microservices effectively, you need to track specific performance and health indicators. Below are nine key metrics commonly used in production environments.

Service Request Tracing

Request tracing records the call chain of a request across multiple services in a microservice architecture.

The typical tracing flow includes:

Request entry: A unique Trace ID is generated and attached to the incoming request.

Service call: The first service processes the request and may call downstream services, passing the Trace ID along.

Record information: Each service logs its processing details together with the Trace ID (e.g., latency, errors).

Pass Trace ID: When invoking another service, the Trace ID is forwarded to maintain the chain.

Request exit: After the final response is returned to the client, the complete trace is aggregated and stored.

Tracing enables performance analysis and fault isolation using tools such as SkyWalking, Zipkin, Sleuth, Jaeger, or PinPoint.

Service Instance Health Status

Health of a service instance is usually monitored through several mechanisms:

Heartbeat Check: Instances periodically send heartbeat signals to a registry or health‑check component. Missing heartbeats indicate the instance may be down.

Health Check: Instances perform self‑checks of critical metrics and dependencies. Failure removes the instance from service discovery.

Load‑Balancer Health Check: Load balancers verify instance liveness; unhealthy instances stop receiving traffic.

Log and Metric Monitoring: Abnormal logs or error‑rate spikes help identify unhealthy instances.

Self‑Healing & Auto‑Scaling: Detected failures can trigger automatic restarts or scaling of new instances.

Throughput

Throughput measures the number of requests a service handles over a period, typically expressed as Requests Per Second (RPS).

For example, a service with a throughput of 100 RPS can process 100 requests each second, indicating its capacity under load.

Request Response Time

Response time measures the latency from receiving a request to sending the response. Shorter times generally reflect better performance (e.g., average 100 ms).

Request Success Rate

Success rate is the proportion of successfully processed requests out of the total, reflecting service availability.

Successful requests: 1,000

Total requests: 1,100

Success rate: 90.91 %

Error Rate

Error rate indicates the fraction of requests that resulted in errors.

Error requests: 50

Total requests: 1,100

Error rate: 4.55 %

Concurrent Connections

Concurrent connections represent the number of client connections active at a given moment.

For instance, 100 concurrent connections mean 100 clients are communicating with the service simultaneously.

CPU and Memory Usage

CPU and memory utilization indicate how much of the instance’s resources are being consumed.

Typical values might be CPU usage 60 % and memory usage 70 %.

Resource Utilization

Resource utilization covers other consumables such as database connection pools, helping assess whether resources are over‑ or under‑used.

These metrics should be selected based on specific business needs and system characteristics.