Monitoring custom metrics; or, How I learned to instrument first and ask questions later
Maxime Petazzoni explains why monitoring custom application metrics is essential for visibility into the internal workings of a system and shares a framework for properly instrumenting them, along with a number of relevant use cases.
| Talk Title | Monitoring custom metrics; or, How I learned to instrument first and ask questions later |
| Speakers | Maxime Petazzoni (SignalFx) |
| Conference | O’Reilly Velocity Conference |
| Conf Tag | Build systems that drive business |
| Location | London, United Kingdom |
| Date | October 31-November 2, 2018 |
| URL | Talk Page |
| Slides | Talk Slides |
| Video | |
What gets measured gets managed, and with the vast array of monitoring systems used today, collecting and alerting on traditional infrastructure metrics is a relatively straightforward task. However, measuring only system metrics or metrics emitted by third-party services won’t paint the whole picture of application health. Custom metrics tend to be underutilized because of the need to modify application code to emit metrics, the higher volumes of data that custom metrics generate, and the challenge of keeping instrumentation consistent across teams as new services are added or old ones changed. But instrumentation friction isn’t as high as it seems; instrumenting custom metrics is critical to correctly monitoring services, as well as for debugging, experimentation, and business observability. Maxime Petazzoni explains why monitoring custom application metrics is essential for visibility into the internal workings of a system and shares a framework for properly instrumenting them while writing application code, along with a number of relevant use cases.
