Dok Talks #115 - What More Can I Learn From My OpenTelemetry Traces? // John Pruitt

https://go.dok.community/slack
https://dok.community/

ABSTRACT OF THE TALK

Of the three observability data types supported by OpenTelemetry (metrics, logs, and traces) the latter is the one with most potential. Tracing gives users insights into how requests are processed by microservices in a modern, cloud-native architecture.

Jaeger and Grafana can visualize a single trace, showing how an individual request traversed your entire system. This helps for distributed debugging and analysis, but using traces only this way is limiting.

What if you stored tracing data in a SQL database? You could ask global questions about your system. You could find slow communication paths, where the error rate spiked since the last deployment, or where the request rate suddenly dropped. Thus, tracing can be used proactively to help you spot issues before your customers do.

This talk will show you how to do all the above by ingesting OpenTelemetry traces into a PostgreSQL/TimescaleDB database, and building custom dashboards using SQL to make the most out of your tracing data.

BIO

John Pruitt is a software engineer at Timescale. His work focuses on database/SQL development for the Promscale open-source observability tool, and currently on adding support for OpenTelemetry tracing. Prior to joining Timescale, John grew the DBA team at Shipt. Most of the balance of his career was spent building custom time-series applications in the energy industry and leading data warehousing efforts at regional banks.

KEY TAKE-AWAYS FROM THE TALK

- What is distributed tracing
- Why viewing individual traces is of limited value
- How SQL can be used to analyze and visualize traces
- What insights can be unlocked using SQL against traces