Observability for Placement Decisions
A shared event contract makes placement decisions explainable across proxies, queues, workers, runtimes, and artifacts without centralizing sensitive workload content.
Notes
28 posts, 2012–2026: reliability engineering, platform architecture, migration constraints, and AI-assisted engineering.
Recent
A shared event contract makes placement decisions explainable across proxies, queues, workers, runtimes, and artifacts without centralizing sensitive workload content.
I write postmortems to explain how a failure reached production, what made the impact worse, and which changes will make the next incident easier to prevent or recover from.
A target can be eligible and still not be ready. Admission needs a bounded capacity claim before competing workloads can start.
From the archive
Earlier writing, kept online for reference. The CentOS, Git deployment, and Linux migration pieces are dated but the approach still holds up; the personal essays from 2012–2014 don't reflect how I write or think now.
2026
2015
2014
2012