Keeping Kubernetes in Check with Prometheus πŸ“Š

Ready to decode metrics? πŸͺ„

Β·

14 min read

Keeping Kubernetes in Check with Prometheus πŸ“Š

πŸ’‘
Check out this article Prometheus Step-by-step Guide if you're new to the world of Prometheus!

Introduction: Charting Your Course

πŸ’‘Welcome to a journey of seamless Kubernetes monitoring powered by Prometheus. Set Prometheus's watchful eye inside your cluster, capturing critical metrics for a profound understanding of your system's health. πŸ“Š

Deploying Prometheus within your Kubernetes cluster offers a powerful vantage point for monitoring your applications and infrastructure. By positioning Prometheus close to your targets, you ensure efficient data collection and real-time insights into performance and resource utilization. πŸš€

Monitoring Kubernetes Cluster includes:

  1. Control Plane Components:

  2. kube-state-metrics:

  3. kubelet (cAdvisor):

  4. Push Gateway:

  5. Node Exporter:


Chapter 1: Helm - Your Trusty Navigator

🎩 Welcome to the Helm helm! In this section, we'll walk you through installing Helm, demystify what a chart is (hint: it's like a mix of value and manifest files), and then get hands-on with deploying Prometheus using Helm charts. Think of it as your GPS for smooth sailing in the world of deployments.

  • Installing Helm: Your Deployment Companion

    🌐 Getting Helm Onboard: Let's start by getting Helm installed. It's like adding a trusty sidekick to your deployment adventures. Just paste these commands, and you're good to go:

      curl -fsSL -o get_helm.sh https://raw.githubusercontent.com/helm/helm/master/scripts/get-helm-3
      chmod 700 get_helm.sh
      ./get_helm.sh
    

    With Helm by your side, you're all set to navigate the seas of deployment.

  • Understanding Charts: The Magic Blueprints

    πŸ’‘ What's a Chart, Anyway?:

    Think of charts as the magical blueprints for your deployment. They bring together value files (for custom settings) and manifest files (to describe your application's needs) in a neat package.

  • Adding Prometheus Chart Repository: Your Beacon

    πŸ—ΊοΈ Chart Repositories: Picture chart repositories as your guiding beacons to find what you need. For Prometheus, we'll tap into its official repository. Just a couple of commands, and you're connected:

      helm repo add prometheus-community https://prometheus-community.github.io/helm-charts
      helm repo update
    

    Now, let's set our course towards deploying Prometheus.

  • Prometheus Deployment: Let's Get Started

    βš“ Setting Sail with Prometheus: This is where Helm's magic shines. Deploying Prometheus becomes a breeze with Helm. Ready? Just type this command and watch the magic happen:

      helm install [RELEASE_NAME] prometheus-community/kube-prometheus-stack
      # [RELEASE_NAME] -- Can be anything of your choice. Use 'prometheus'
    

    By default, this chart installs additional, dependent charts:

Like a captain summoning the crew, Helm takes care of the rest and brings Prometheus to life.

πŸ”š For uninstalling a helm chart :

    helm uninstall [RELEASE_NAME]
  • Fine-Tuning the Journey: Customization Made Easy

    πŸ› οΈ Fine-Tuning with Value Files: Helm's your customization wizard. To tweak Prometheus' journey, use value files to adjust settings to your liking:

      # check values.yaml for all the settings to customize and install
      helm show values prometheus-community/kube-prometheus-stack > values.yaml
      helm install prometheus prometheus-community/kube-prometheus-stack -f values.yaml
    

    Much like tailoring a suit, you're shaping Prometheus' course to fit perfectly into your Kubernetes world.

  • Helm Upgrade: Elevate and Adapt

    πŸ“ˆ Chart Evolution with Helm: Helm's prowess extends beyond the initial deployment. It's your compass for upgrades. When a newer version of Prometheus beckons, Helm enables seamless upgrades:

      helm upgrade prometheus prometheus-community/kube-prometheus-stack
    

Chapter 2: Prometheus Chart - The North Star of Metrics

πŸ”In this segment, we'll take a deep dive into the components that compose this remarkable chart.

kubectl get all
  • StatefulSets: The Backbone of Persistent Services
    🏰 StatefulSets orchestrate the deployment of stateful applications. In the context of the chart, they bring forth the powerful Prometheus servers and Alertmanager instances. These StatefulSets ensure that Prometheus' vital data is persistently stored and managed, guaranteeing resilience and accuracy in your monitoring.

  • Deployments: Scaling with Grace
    πŸš€ Scaling Through Deployments: Deployments handle the rollout and scaling of stateless applications. Within the chart, they manage components like the Grafana, Prometheus Operator and kube-state-metrics. Join us as we delve into the unique role, significance, and collective impact of these components on your Kubernetes monitoring ecosystem.

  • DaemonSets: Agents on Every Node
    🚀 Exploring DaemonSets: DaemonSets ensure that a specific pod runs on every node in your cluster. The chart's DaemonSets, exemplified by the Node Exporter, provide critical node-level metrics. These agents ensure comprehensive monitoring coverage by collecting data from every node, enhancing your understanding of cluster health.

  • Pods: Collaboration in Containers
    πŸ“¦ Pods in Harmony: Pods encapsulate one or more containers, sharing the same network space. Within the chart, pods house vital components like Prometheus, Grafana, and more. These pods collaborate to create a cohesive monitoring environment, allowing different components to work together seamlessly.

  • Services: Networking and Connectivity
    🌐 Navigating Services: Services expose pods to the network, enabling communication within and beyond the cluster. In the chart, services connect components like Prometheus and Grafana to facilitate interactions. They play a pivotal role in ensuring seamless data flow and access across your monitoring ecosystem.

  • ReplicaSets: Ensuring Availability and Resilience
    πŸ” ReplicaSets in Action: ReplicaSets guarantee that a specified number of replicas are available at all times. While not directly managed in the chart, they contribute to the reliability of your monitoring components. ReplicaSets ensure that your Prometheus, Alertmanager, and other components maintain the desired level of availability.

Seamless Rule Configuration: Accessing Rule Configurations from Prometheus StatefulSet

  • Step 1: Identifying Your StatefulSet Prometheus Instance
    πŸ” Begin by identifying the StatefulSet-managed Prometheus instance you want to customize. You can list your running StatefulSets using:

      kubectl get statefulsets -n <namespace>
    

    Locate the Prometheus StatefulSet you want to customize.

  • Step 2: Accessing the Configuration Map

    πŸ“‚ Let's access the configuration map that holds your rule files. Use the following command to describe the StatefulSet and find the name of the configuration map:

      kubectl describe statefulset <prometheus-statefulset-name> -n <namespace> > prometheus-instance.yaml
    

    Look for the "ConfigMap" section. This is where you'll find the name of the configuration map that holds your rule files.

  • Step 3: Describing the Configuration Map
    πŸ“‹ Now, let's get a closer look at the configuration map. Run the following command:

      kubectl describe configmap <configmap-name> -n <namespace>
    

    This will provide detailed information about the configuration map, including the rule files it contains.

  • Step 4: Operator's Toolkit in the Deployment Set

    πŸ”§ As we journey through the Deployment set, remember the Operator's toolkit. Within this set, the Prometheus Operator orchestrates the deployment, management, and configuration of Prometheus instances. It's here that the magic happensβ€”grabbing the configuration map for rule files directly from the Prometheus instance managed by the Operator.

      kubectl describe deployment <prometheus-operator> -n <namespace>
    

Chapter 3: Connecting to Prometheus - Anchoring Your Insights

In this section, we'll dive into the various methods to connect to your Prometheus instance and interact with the metrics it collects. Prometheus offers powerful insights, but how you access it greatly influences your monitoring experience.

  • Step 1: Initial Access using ClusterIP

    By default, Prometheus often uses a ClusterIP service type, limiting access to within the cluster. To connect to Prometheus using this setup, you can use the internal ClusterIP address. Keep in mind that this access is limited to within the cluster environment.

  • Step 2: Upgrading to NodePort or LoadBalancer

    For external access to Prometheus, consider transitioning to NodePort or LoadBalancer service types. Here's how you can upgrade to NodePort:

      kubectl edit svc <prometheus-service-name> -n <namespace>
      # Change type: ClusterIP to type: NodePort
    

    For LoadBalancer:

      kubectl edit svc <prometheus-service-name> -n <namespace>
      # Change type: ClusterIP to type: LoadBalancer
    
  • Step 3: Using Port-Forward for Temporary Access

    For quick and temporary access during development, use the kubectl port-forward command:

      kubectl port-forward <prometheus-pod-name> 9090:9090 -n <namespace>
    

    Access Prometheus at http://localhost:9090 in your browser.

  • Step 4: Accessing Prometheus via Localhost:9090

    After port-forwarding, open your browser and navigate to http://localhost:9090. You'll land on the Prometheus web interface, ready to execute queries, explore metrics, and glean insights from your monitored resources.


Chapter 4: Prometheus Configuration

In this section, we'll guide you through configuring Prometheus to harness the power of Kubernetes service discovery. By automatically discovering and monitoring Kubernetes resources, you'll lay the foundation for effective monitoring.

  • Step 1: Embracing Kubernetes Service Discovery Configurations
    Prometheus offers tailored configurations for Kubernetes service discovery. These configurations enable Prometheus to dynamically detect and monitor your Kubernetes pods and services, adapting seamlessly to changes in your environment.

  • Step 2: Exploring Prometheus Web UI: Status > Configuration

    1. Open your browser and access the Prometheus web interface.

    2. Click on the "Status" tab, followed by "Configuration."

  • Step 3: Unveiling Kubernetes Service Discovery Details

    1. In the "kubernetes-sd-configs" section, explore the breakdown of your Kubernetes service discovery configurations.

    2. Look for the category labelled "role: endpoints". This is where you uncover the "endpoints" category, revealing the endpoints associated with your Kubernetes service discovery.

  • Step 4: Checking Targets in the Service Discovery Tab

    1. Navigate to the "Service Discovery" tab within the Prometheus web UI.

    2. Here, you'll find a list of targets that Prometheus has successfully discovered and connected to.

    3. The "UP" status beside each target indicates its accessibility and readiness for monitoring.


Chapter 5: Deploying the Demo App

In this section, we'll guide you through deploying a simple Flask demo application onto your Kubernetes cluster.

  • Step 1: Develop a Dummy Flask App

    Create a directory named "demo-flask-app" and within it, create a file named app.py with the following content:

      from flask import Flask
    
      app = Flask(__name__)
    
      @app.route('/')
      def hello():
          return "Hello, Kubernetes Flask App!"
    
      if __name__ == '__main__':
          app.run(host='0.0.0.0')
    
  • Step 2: Crafting a Dockerfile

    In the "demo-flask-app" directory, create a file named "Dockerfile" with the following content:

      FROM python:3.9
      WORKDIR /app
      COPY requirements.txt .
      RUN pip install -r requirements.txt
    
      COPY . .
      CMD ["python", "app.py"]
    

    Create a file named "requirements.txt" with the following content:

      Flask==2.0.1
    
  • Step 3: Building the Docker Image

    Navigate to the "demo-flask-app" directory and run the following command to build the Docker image:

      docker build -t demo-flask-app:v1 .
    
  • Step 4: Creating Kubernetes Deployment and Service Manifests
    Now, let's deploy the simple flask application with 3 replicas which listen and expose metrics on port 8080.

    Create a file named "demo-flask-app-deployment.yaml" with the following content:

      apiVersion: apps/v1
      kind: Deployment
      metadata:
        name: example-app
      spec:
        replicas: 3
        selector:
          matchLabels:
            app: example-app
        template:
          metadata:
            labels:
              app: example-app
          spec:
            containers:
            - name: example-app
              image: demo-flask-app:v1
              ports:
              - name: web
                containerPort: 8080
    
      ---
      apiVersion: v1
      kind: Service
      apiVersion: v1
      metadata:
        name: example-app
        labels:
          app: example-app
      spec:
        selector:
          app: example-app
        ports:
        - name: web
          port: 8080
    

    Let's expose the application with a Service object which selects all the Pods with the app label having the example-app value. The Service object also specifies the port on which the metrics are exposed.

  • Step 5: Deploying the Demo Flask App
    Apply the deployment and service manifests using the following commands:

      kubectl apply -f demo-flask-app-deployment.yaml
    
  • Step 6: Accessing the Demo Flask App

    Retrieve the assigned NodePort using the following command:

      kubectl get svc demo-flask-app-service
    

    In your browser, navigate to http://<Node-IP>:<NodePort> to access the demo Flask app.


Chapter 6: Service Monitors - Navigating Specifics

In this section, we'll delve into the world of Prometheus Custom Resource Definitions (CRDs), specifically focusing on Service Monitors. As part of the Prometheus Operator's toolkit, Service Monitors play a pivotal role in automating the discovery of targets for Prometheus.

  • Understanding CRDs: Custom Resource Definitions

    Custom Resource Definitions (CRDs) extend Kubernetes to incorporate your own resource types. They enable you to define and manage resources beyond the built-in ones like pods and services. Prometheus Operator leverages CRDs to bring Kubernetes-native declarative configuration to Prometheus deployments.

  • Exploring Service Monitors: Automating Target Discovery
    Service Monitors are a prime example of Prometheus Operator's CRDs. They define how Prometheus should monitor a specific service. By specifying labels that match your target services, Service Monitors automate the discovery of endpoints to be monitored. This dynamic setup adapts as your services scale or change.

  • Connecting Service Monitors to Demo Flask App
    Imagine our "demo-flask-app" as a service you wish to monitor. Let's create a Service Monitor "demo-flask-app-servicemonitor.yaml" to automate its discovery for Prometheus:

      apiVersion: monitoring.coreos.com/v1
      kind: ServiceMonitor
      metadata:
        name: example-app
        labels:
          team: frontend
      spec:
        selector:
          matchLabels:
            app: example-app
        endpoints:
        - port: web
    

    Here, we create a ServiceMonitor object which selects all Service objects with the app: example-app label. The ServiceMonitor object also has a team label (in this case team: frontend) to identify which team is responsible for monitoring the application/service.

  • Apply the Service Monitor to your Kubernetes cluster
    Run the following command to deploy the Service Monitor:

      kubectl apply -f demo-flask-app-servicemonitor.yaml
    

    This will create the Service Monitor object in your cluster, and Prometheus Operator will pick it up and configure Prometheus to scrape the specified targets according to the configuration.

    Remember to adjust the labels and other configurations to match your specific application and monitoring requirements.

    Let's now proceed to the Prometheus web UI and verify the status of our monitoring targets. This step is crucial to ensure that we are indeed receiving metrics from our application.


Chapter 7: Adding Rules - Crafting Alerts Like a Pro

In this section, we'll dive into the concept of adding custom rules to your Prometheus monitoring setup using the PrometheusRule Custom Resource Definition (CRD). By defining these rules, you can fine-tune alerting thresholds and other behaviours to match your specific application and infrastructure needs.

  • Introducing PrometheusRule CRD: Crafting Custom Monitoring Rules

    The PrometheusRule CRD provides a powerful way to extend and customize your monitoring capabilities within the Prometheus ecosystem. With this CRD, you can define alerting and recording rules that suit your application's unique characteristics. These rules govern when alerts should be triggered and what metrics should be recorded, giving you greater control over your monitoring precision.

  • Example: Custom Alerting Rule
    Let's consider an example "custom-alert-rules.yaml" where you want to set up a custom alerting rule for high CPU usage in a particular application. Here's how you can create a PrometheusRule for this purpose:

      apiVersion: monitoring.coreos.com/v1
      kind: PrometheusRule
      metadata:
        name: custom-alert-rules
        labels:
          release: prometheus
      spec:
        groups:
          - name: high-cpu-alerts
            rules:
              - alert: HighCPULoad
                expr: node_cpu_seconds_total / node_memory_MemTotal * 100 > 80
                for: 5m
                labels:
                  severity: critical
                annotations:
                  summary: High CPU Usage Detected
                  description: High CPU usage detected on {{ $labels.instance }}
    
  • Apply the PrometheusRule to your cluster
    Run the following command to apply the rule to your Kubernetes cluster:

      kubectl apply -f custom--alert-rules.yaml
    
  • Verify the rule's status

    You can check the status of the applied rule using the following command:

      kubectl describe prometheusrule custom-alert-rules
    

    This command will provide detailed information about the PrometheusRule, including its groups, rules, and other configuration details.


Chapter 8: Alertmanager Rules - Keeping a Weather Eye Open

In this section, we'll delve into the realm of Alertmanager rules, which play a pivotal role in defining how alerts are managed, grouped, and dispatched. By harnessing Alertmanager rules, you gain the power to shape your alert notifications in ways that align with your operational priorities.

  • Decoding Alertmanager Rules
    Alertmanager rules are your toolkit for sculpting the behaviour of alerts, enabling you to control their routing, deduplication, grouping, and notification channels. With these rules, you shape how alerts are handled, ensuring that critical incidents receive the attention they deserve while minimizing alert fatigue.

  • Scenario: Differentiating Critical and Minor Alerts
    Consider a scenario where you want to differentiate between critical and minor alerts. Here's how you can achieve this using Alertmanager rules.

    Save the following content as alertmanager-custom-rules.yaml:

      apiVersion: monitoring.coreos.com/v1
      kind: AlertmanagerConfig
      metadata:
        name: alertmanager-custom-rules
        labels:
          release: prometheus
      spec:
        route:
          routes:
            - match:
                name: severity
                value: critical
              receiver: critical-email
            - match:
                name: severity
                value: minor
              receiver: minor-email
            - match:
                name: severity
                value: minor
    
      receiver: slack-notification
        receivers:
          - name: critical-email
            email_configs:
              - to: critical@example.com
                send_resolved: true
          - name: minor-email
            email_configs:
              - to: minor@example.com
                send_resolved: true
          - name: slack-notification
            slack_configs:
              - api_url: https://hooks.slack.com/services/your/slack/webhook/url
                channel: "#alerts"
                text: "Alert: {{ .CommonAnnotations.summary }}"
    

    For critical-email and minor-email receivers, email_configs are defined to send alerts to the specified email addresses. The send_resolved field is set to true to also send "resolved" alerts.

    For the slack-notification receiver, slack_configs are defined to send alerts to a Slack channel. You'll need to replace api_url with your actual Slack webhook URL.

Updating the alertmanagerConfigSelector

  1. Locate Helm Values for the Prometheus Chart

     helm show values prometheus-community/kube-prometheus-stack > values.yaml
    

    This command fetches the default values for the prometheus-community/kube-prometheus-stack chart and saves them to a file named prometheus-values.yaml.

  2. Edit Helm Values

     alertmanagerConfigSelector:
       matchLabels:
         release: prometheus
    

    This configuration tells Prometheus to use your custom Alertmanager configuration associated with the prometheus release.

  3. Upgrade the chart

     helm upgrade prometheus prometheus-community/kube-prometheus-stack -f values.yaml
    

    This command upgrades the prometheus release with the modified values, ensuring that it now references your custom Alertmanager configuration.

πŸ’‘Why We're Doing This:

By updating the alertmanagerConfigSelector in the Helm values, you're instructing Prometheus to associate your custom Alertmanager configuration with the Prometheus deployment. This ensures that the alerting rules and notification configurations you've defined will be correctly integrated and utilized by the Prometheus Operator.

This integration is essential for maintaining consistency between your monitoring components and making sure that your custom configurations take effect as expected.


Conclusion: Setting Sail for Prometheus Glory

β›΅ Setting Sail for Prometheus Glory: Bravo, fellow mariner of metrics! You've conquered Prometheus like a true captain. With these newfound skills, you're poised to navigate Kubernetes waters with confidence and humour. Remember, the journey doesn't end here. Keep exploring, keep monitoring, and may your data always flow as smoothly as a calm sea.

See you on the next one. Cheers! βœ¨πŸš€


Β