LogicMonitor recently released LM Container Argus v4 and has made monitoring Kubernetes clusters even easier. We have updated our Add Kubernetes Cluster wizard to make customization easier in the installation process. Start by going through the normal configuration settings, and then you will have the option to edit additional configuration settings to the CollectorSet Controller and Argus YAML files. These additional configuration settings include parameters for discovery filtering and PriorityClasses, to name a few. For users that want more collector replications or have their own collector image in a private repository, they can make the modifications in the GUI via the text editor.
By default, these settings are blank and you can add configuration as needed.
Focus on What’s Important
Discovery filtering, similar to what is available via LM Cloud, can exclude certain resources from monitoring based on labels. You have more flexibility to exclude resources from unused or low usage development environments. For larger clusters where API server resources are not scalable, discovery filtering can help you better manage API server load.
Discovery filtering can be configured in the LogicMonitor UI (from the Add Kubernetes Cluster wizard) or directly in your Helm deployment.
Another example of how tags are useful; a user may tag certain environments that correspond to blue-green deployments and only monitor the active one. They can also tag pods that correspond to CICD processes that are constantly spinning up and down and don’t require monitoring due to their short lifespan. Also, Kubernetes comes with infrastructure that may not provide much monitoring value like Prometheus, Fluentd, etc. Thus, filtering these devices can save valuable resources in the API server.
Keep Your Space Up to Date
Another important improvement is that Argus will now retry adding or deleting resources in monitoring when the initial request fails. This compensates for any issues that may interfere with the initial request, ensuring monitoring stays up to date. Along with this, LogicMonitor can now handle pods with the same name that are running in different namespaces without any issues. The pod naming convention changes from <pod name> to <pod name-namespace>.
Kubernetes users running v1.14+ have the ability to create more sophisticated scheduling policies to better manage the workload of their pods through the use of PriorityClasses. LogicMonitor has added support of PriorityClasses for Argus, Collectorset-controller, and collector pods. Users can now assign a priority to these components to have greater control over their resource utilization.