Kiota OpenTelemetry PoC

[calebkiage]

Concept to prove: using the new open telemetry Azure AppInsights SDK will give us the data we want based on the telemetry spec.

In scope:

• Export metrics & traces to a temporary azure app insights instance.
• Setting up a query on AppInsights to answer the question What are the most used commands?.

Out of scope

• Exporting logs
• Opt-out functionality

TODO

• Set up OpenTelemetry metrics e.g. command count, language generation count
• Set up OpenTelemetry traces e.g. command execution trace
• Add a temporary Azure ApplicationInsights instance for PoC metrics submission
• Send data to ApplicationInsights instance

Misc tasks

• Clean up labels for the different OTel instruments and attributes for better queries
• Create a query for one scenario in the spec (What are the most used commands?)

Post-Poc (once labels are configured)

• Switch the exporter to the Azure Application insights live instance.
• Create other queries to give full view of data from spec

[maisarissi]

Hey @calebkiage, based on our discussion today, please also add as part of the post-poc the task of figuring out where/how to store/persist the data.

1. We will need to persist or log the data to see trends/adoption based on the questions we want to answer.
2. The data can't only live in app insight because we need the data over time. Also, if the data lives in the azure app insights, this would require a SAW machine and Torus account to access it, and this data needs to be easily shareable with other teams/orgs.
3. We can keep tracing with engineering metrics as part of the app insight to help us analyze Kiota, that doesn't need to be persist elsewhere nor be available in the dashboard.
4. After all data we need to answer the questions in the telemetry spec are collected, I'll work with the data team and build a dashboard on top of the collected data.

Thanks.