At scale, observation of metrics becomes important, to be able to feed it to the service dashboard that can alert engineers to problems. Metrics of CPU/Mem/etc are great, more added value could be via cadvisor/node-exporter type metrics that are currently in the cloud native landscape. (especially with microservice arch on devices)
Ken Bannister: See our experimental OpenTelemetry Collector repository and blocks. It uses node-exporter to collect metrics.
Ken Bannister: Published a blog post on the OpenTelemetry Collector investigation and block.
Ken Bannister: We just updated the experimental collector container to the latest upstream, and added device log collection. Let us know how that works for you.