tag
6 posts
How to identify high-cardinality metrics and labels, stop unsafe ingestion, preserve evidence, and redesign instrumentation without masking outages.
How tail sampling buffers complete traces, evaluates outcome-aware policies, controls memory, and preserves representative telemetry during failures.
How to design Collector pipelines for metrics, logs, and traces without creating an unbounded queue, telemetry leak, or new single point of failure.
Prometheus became the CNCF's second hosted project in 2016, placing metrics and alerting beside orchestration at the center of cloud native systems.
How the three pillars of observability complement each other, and why having all three matters more than maximizing any single one.
Understand eBPF programs, verifier, maps, hooks, BTF and CO-RE, privileges and operational risks without treating kernel execution as a sandbox.