Question: Is it possible to view CPU/RAM usage of remote sessions in the administrative dashboard of RStudio Workbench (previously RStudio Server Pro) when using Launcher and Kubernetes?
Answer:
Yes, you can view CPU/RAM usage of remote sessions in the administrative dashboard of RStudio Workbench (previously RStudio Server Pro) when using Launcher and Kubernetes.
To enable this functionality, the Launcher Administrator Guide mentions the following requirement:
The cluster must have the metrics-server addon running and working properly to provide job resource utilization streaming
The metrics-server is a "cluster-wide aggregator of resource usage data" and is one of many third-party projects available for monitoring resource usage on Kubernetes clusters.
You can install the third-party metrics-server on your Kubernetes cluster by following the documentation on the metrics-server page. After you install the metrics-server on your Kubernetes cluster, then you should see the CPU and RAM columns populate in the "Sessions" tab in the administrative dashboard of RStudio Workbench for remote sessions running in Kubernetes.
Note that the CPU/RAM information will only be shown in the "Sessions" tab in the administrative dashboard of RStudio Workbench, but the Kubernetes cluster usage will not be visible on the "System Utilization" tab or included in the RRD monitoring data written from RStudio Workbench.
Question: Are there other options for monitoring and recording usage metrics of RStudio Workbench when it's configured to use Launcher with Kubernetes?
Answer:
For monitoring resources in Kubernetes (outside of administrative dashboard in RStudio Workbench), there are two recommended approaches, both of which are outlined in the Kubernetes Tools for Monitoring Resources.
If you use the recommended option for a "Resource metrics pipeline" using the metrics-server deployed in Kubernetes, then RStudio Workbench (configured with Launcher and Kubernetes) will display the CPU/RAM metrics for remote RStudio sessions in Kubernetes, but this information will not be included in the RRD monitoring data written from RStudio Workbench.
If you use the recommended option for "Full metrics pipeline" outside of RStudio Workbench, you can get more detailed metrics on the containers and workloads on the Kubernetes cluster using tools such as Prometheus.
Comments