Virtualization Technology News and Information
Article
RSS
4 Monitoring Tips for Container Mastery

By Sarah Terry, Manager, Product Management at LogicMonitor

According to a Gartner survey, more than 50% of global organizations will be running containerized applications in production by 2020, up from less than 20% in 2017. This shift is understandable, as containers enable greater velocity and efficiency when developing, deploying and scaling applications. However, these advantages also come with additional challenges for enterprise IT teams. Comprehensive container monitoring can ensure that organizations maximize their container investments and overcome common container pitfalls. Here are four tips to get started:

1. Improve visibility with event-based discovery

Continuous, comprehensive visibility into your containerized workloads is necessary to proactively prevent downtime. It seems obvious, but with the ephemeral nature of containers, this visibility can be challenging to achieve because the resources that need to be monitored are constantly changing.

Discovery mechanisms that rely on periodically making requests to check for new resources likely won't be fast enough to capture ephemeral containers in a timely manner. They're also inefficient.

To avoid this inefficiency and ensure full visibility, use a monitoring solution with built-in automation. This should include event-based discovery for containers, such that your monitoring accurately reflects your running containers at all times. With event-based discovery, the occurrence of an event will automatically trigger an update to what is in monitoring.

2. Prevent downtime & boost efficiency with orchestration monitoring

Container orchestration tools like Kubernetes enforce a declared state for your containerized workloads and can make it easier to manage your applications. Because the availability of your workloads relies on orchestration, you should monitor your container orchestration (e.g. Kubernetes) in addition to your containers and applications.

Additionally, monitoring your container orchestration can help you identify where containerized applications can be optimized. For example, Kubernetes has resource requests and limits that designate minimum and maximum levels of resource consumption for a specific application. These requests and limits determine how Pod workloads are scheduled across Nodes. By monitoring requests and limits alongside application CPU and memory consumption, you can identify applications that are reserving more resources than they need.

When considering solutions, select a monitoring platform that is capable of automatically discovering and monitoring your container orchestration, such as Kubernetes cluster resources, and alerting you when they are not in good health. Double check that the platform provides visibility into the Pods, Containers, the worker Nodes in the cluster, and the cluster's master components responsible for orchestration.

3. Save time with automation

To maximize visibility into containerized applications, you should minimize the configuration needed to start data collection and alerting. For example, setting alert thresholds manually or editing a configuration file to start data collection will delay how soon you see metrics and receive alerts. Given the average lifetime of containers, this delay can make a big difference in your ability to troubleshoot issues.

Ideally running applications are automatically detected right away, so that appropriate monitoring and alerting can begin. Automatic detection of a running application usually depends on the application's accessibility. Container networking can make accessibility challenging, and therefore can make it harder to monitor applications via traditional methods.

With default Docker networking, containers can only talk to one another if they are on the same server, and any external exposure requires a unique port on the server. This requires you to install a monitoring agent or Collector on every server (ideally containerized, such that you can use the same ports across containers). Orchestration tools like Kubernetes solve a lot of these challenges with cluster networking, in part by providing each Pod in the cluster with a unique IP, enabling more traditional and agentless methods for monitoring containerized applications.

Look for a monitoring tool that comes with out-of-the-box support for monitoring common containerized applications. A tool that can automatically detect running applications by making requests to common ports and that has pre-configured standards for data collection will require the least end-user configuration. It's also a good idea to select a monitoring solution that has the flexibility to customize how data is collected.

4. Ensure app health through aggregation

Aggregate data across containers running the same application in order to get an accurate picture of the overall health of the application. Intelligent IT monitoring software enables you to aggregate data dynamically based on tags (e.g. Kubernetes labels assigned to Pods all running the same application). Focusing on the underlying containers may be necessary for troubleshooting, but won't provide you with long-term insight into your application's performance. Aggregation, on the other hand, allows comparisons that provide insights into if app performance has improved or degraded over time, or if a recent release impacted performance.

##

About the Author

Sarah Terry 

Sarah Terry is a long time tech enthusiast. She has dabbled in many interesting projects including researching lasers at CalTech. As a manager in product management at LogicMonitor, Sarah works closely with customers, defines product requirements, collaborates with R&D, and helps deliver new functionality to customers. Her goal is to ensure that SaaS monitoring continues to enable customers to innovate faster and manage their businesses more efficiently. When she is not working, you will find Sarah riding her horse, doing barre classes, or rock climbing.

Published Thursday, October 31, 2019 7:36 AM by David Marshall
Comments
There are no comments for this post.
To post a comment, you must be a registered user. Registration is free and easy! Sign up now!
top25
Calendar
<October 2019>
SuMoTuWeThFrSa
293012345
6789101112
13141516171819
20212223242526
272829303112
3456789