The Complete Guide to Container Monitoring

The benefits for deploying containers are obvious. At the same time, the natural behavior of containers requires a different approach to monitoring than physical hardware. While retrieving monitoring data, monitoring status, and setting thresholds are the same for both types of deployments, that's about it. Containers are ephemeral and can have a lifespan of only a few seconds, making it difficult to track changes. Containers also share resources, making it difficult to know the health of an application and the performance of containers.

The main challenges in monitoring containers are:

• Tracking containers throughout their lifecycle.
• Scaling monitoring as the container environment grows.
• Checking the resource usage of individual containers and shared resources such as storage.
• Defining reasonable thresholds for alerting on problems.

To overcome these challenges, container monitoring requires solutions that have the following features, among others:

• Automatic detection to discover new containers and add them to monitoring.
• Automatic inventory of all hardware and software aspects of nodes and containers.
• Simple scalability of the monitoring.
• Recording of performance metrics such as CPU utilization, memory, or network transfer rates of the individual containers.
• Visual representation of container metrics for easier problem detection.
• Support for orchestration tools such as Kubernetes or Docker Swarm or cloud platforms on which containers are hosted.
   

There are two ways to monitor containers. One is to use the command line of orchestration platforms, such as Docker or Kubernetes, to access logs and information about the containers. This gives you a basic overview, and you can use the logs to analyze possible causes of errors. However, the main drawback is that manual monitoring is inefficient and time-consuming, and it's easy to lose track. In addition, this method does not provide alerts when problems occur.

The second option is to use a monitoring solution that is designed specifically for effective container monitoring and that provides all the necessary visibility into container resource usage along with features such as historical data analysis, data visualization, and alerting.

The Docker command line includes some commands that you can use for monitoring Docker containers:

• The docker stats [OPTIONS] [CONTAINER...] command displays real-time resource usage statistics for all running containers, providing insight into CPU, memory usage, memory limit, and network IO metrics. If you wish to focus on specific container(s), you can specify a list of container names or IDs separated by spaces.
• You can view the running container list using the docker ps [OPTIONS]. Use the -all command to show all containers.
• You can use the docker top CONTAINER [ps OPTIONS] command to view the processes for a specific container.

Alternatively, Docker containers can be monitored with a monitoring solution. The benefits of dedicated container monitoring software include:

• Detailed information about the performance and resource usage of Docker nodes and containers. For example, CPU utilization, memory usage, and disk I/O.
• Visualization of key metrics and time series data in dashboards.
• Contextual information for effective root cause analysis and complex problem-solving.
• Storage of historical data to analyze trends or changes in the Docker topology.
• Alerts on critical events or anomalies in the container environment.

For example, Checkmk uses an agent to obtain detailed information about the Docker node and every single container.

Kubernetes provides the ability to monitor containers through the metrics server. This makes it possible to retrieve a limited set of metrics, such as CPU and memory utilization, via the Kubernetes Metrics API. With the Kubernetes Dashboard, the container platform also offers a web-based user interface that provides a convenient way to manage and monitor the containerized applications in a Kubernetes cluster once deployed:

• The Kubernetes metrics server enables collecting metrics, such as CPU and memory usage, via the Kubernetes Metrics API.
• The Kubernetes Dashboard is a web-based interface that allows you to:
  • Deploy and troubleshoot applications.
  • Manage cluster resources.
  • Monitor containerized applications running and modify Kubernetes resources like deployments, jobs, and daemon sets.
  • Perform actions such as scaling deployments, initiating rolling updates, restarting pods, and deploying new applications.
  • Gain visibility into cluster resources and errors.

Checkmk uses the API server to get basic information about the cluster. The Checkmk Node Collector and the Checkmk Cluster Collector, which provides information about how much memory a DaemonSet is currently consuming, offer a comprehensive picture and detailed insights into container behavior. Integrating a dedicated Kubernetes monitoring software like Checkmk enhances capabilities by:

• Collecting an extensive set of metrics beyond Kubernetes' default monitoring scope.
• Monitoring Kubernetes components – including clusters, nodes, deployments, pods, daemon sets, and stateful sets – while tracking key metrics such as CPU, memory, file systems, disk I/O, kernel performance, and threads.
• Intelligent alerting: Checkmk Kubernetes monitoring provides smart alerts that account for Kubernetes' self-healing features, notifying you only if issues persist beyond a certain threshold.
• A comprehensive view of your Kubernetes environment through Checkmk’s interconnected dashboards, offering valuable contextual information for troubleshooting and high visibility into containers.