The central idea of this talk was a parallel that kept nagging me: if test-driven development changed how we write code, shouldn’t monitoring-driven development change how we run it? TDD gives developers fast feedback during development. Monitoring should give the same tight feedback loop in production. The problem was that in 2012, monitoring was still firmly an ops concern – developers deployed and moved on.
I framed the talk around the four areas of DevOps and the CAMS model. Extending development practices into operations meant applying the same rigor – version control, testing, code review – to monitoring configurations. The feedback direction mattered too: operations data flowing back to developers so they could learn from how their code actually behaved under real load, not just in test environments.
The monitoring tools landscape at the time was shifting fast. Nagios was the incumbent but showing its age. Sensu emerged as a more modern alternative with better extensibility. Collectd handled system-level metrics efficiently. Graphite became the go-to for time-series storage and visualization. And then statsd appeared and changed everything by making it trivially easy for developers to emit custom metrics from application code – one UDP packet, no blocking, no ceremony.
Self-service metrics was the breakthrough pattern. Instead of filing a ticket for ops to add a monitor, developers could instrument their code directly. Logstash brought the Unix pipe philosophy to log processing – small composable pieces that could be chained together. This was infrastructure as code applied to the monitoring stack itself.
The concept I called “meta-ops” tied it all together: annotating your monitoring with change events. Every deployment, every config change, every feature flag toggle becomes a marker on your graphs. When something goes wrong, you don’t have to guess what changed – the correlation is visible. Combined with CI pipelines that tested not just application code but infrastructure and monitoring configurations, this created a genuine feedback loop from production back into development.
Watch on YouTube — available on the jedi4ever channel
This summary was generated using AI based on the auto-generated transcript.