ABOUT THE TALK:
Logs, time series metrics, dashboards -- the tools of the LAMP stack era are mature, robust and even somewhat commoditized. But architecture and deployment patterns have revolutionized many times over these past 10-15 years, and the rate of change is still accelerating. Modern infrastructure is fluid, dynamic, chaotic and massively complex, and logs and metrics are straining to usefully address its problems. The mess of hacky glue code and one-offs everywhere is a big SUV that most of modern "monitoring" is busy churning out better buggies, not engines for our cars.
What problems are metrics and logs most fundamentally ill-equipped to address? What alternatives are there? And what comes next?
* Rated "H" for possible hyperbole, strong language.
* * Rated "Honey" for some discussion of how we are attacking the problem set with honeycomb, but this is NOT a sales pitch and I'll tell you all the ways we fall short too.