4 key metrics to know when monitoring microservices applications running on Kubernetes

Understanding how microservice applications works on Kubernetes is important in software development. In this article, we will discuss why observing microservice applications on Kubernetes is crucial and several metrics that you should focus on as part of your observability strategy.

Why should you observe microservice health running on Kubernetes and what are the Kubernetes metrics you should monitor?

Consider a large e-commerce platform that utilizes microservices architecture deployed on Kubernetes clusters. Each microservice, responsible for specific functionalities such as inventory management, order processing and payment handling, operates independently and communicates with others via APIs which are critical to your business/ service growth.

In such a complex environment, ensuring seamless operation and detecting issues proactively becomes imperative and can be challenging.

Observability in this scenario can assist with real-time insights into the performance, availability and interdependencies of these microservices and the Kubernetes application.

Observability is essential for several reasons:

Early detection of issues: Microservices are distributed and interconnected, making it challenging to identify issues when they arise. Observing their health allows you to detect problems early on, minimizing downtime and potential service disruptions. Using Instana you will get 1-second granularity, which helps you detect problems faster than other solutions.
Reliability: Monitoring microservice health ensures that your application remains reliable. By tracking metrics such as response times, error rates and resource utilization, you can proactively address any performance issues before they impact users.
Scale efficiently: Kubernetes allows for dynamic scaling of microservices based on demand. Observing their health helps you make informed decisions about when and how to scale services to ensure optimal performance and resource utilization.
Meet SLAs: Many organizations have service-level agreements (SLAs) that define expected levels of service availability and performance. Observing microservice health helps you meet these SLAs by ensuring that your services are running smoothly and meeting performance targets.

By monitoring Kubernetes health, organizations can proactively identify and address issues, optimize resource usage and maintain optimal cluster performance. These are the key metrics that can be measured:

Cluster availability:

Monitoring Kubernetes cluster availability metrics helps ensure that the clusters are up and running and are healthy. Metrics such as cluster uptime and pod status provide insights into the overall health of the cluster. They are at the highest and most important layer and can provide complete visibility into what’s happening in your environment.

Pod metrics:

Monitoring pod health metrics such as pod restarts, pod readiness and pod eviction help identify issues with individual pods and ensures that applications are running smoothly. Monitoring pod health enables organizations to detect and troubleshoot issues quickly, minimizing downtime and ensuring high availability.

Service availability:

Monitoring service availability metrics such as service uptime, service response time and service error rate help ensure that Kubernetes services are available and responsive to users. By monitoring service availability, organizations can detect service failures or degradation and take proactive measures to restore service availability and minimize impact on users.

Nodes health:

This is a metric that shows the status of nodes in the context of Kubernetes cluster metrics. Some other important metrics include:

kube_node_status_capacity: This metric indicates the available capacity for different resources on a node, helping you identify how much resources are available.
kubelet_running_container_count: It tells you how many containers are currently running on a node.
kubelet_runtime_operations_latency_microseconds: This metric measures the time it takes for each operation to complete, categorized by type, and it’s measured in microseconds.

Observability by the numbers

IBM Instana can monitor your microservice application running on Kubernetes.

IBM Instana is a fully automated real-time observability platform that contextualizes performance data. It lets you detect problems or transactions with a 1-second granularity in your microservice application. Additionally, you get 100% traces that allow you to fix issues easily if there are any when running your microservices on Kubernetes.