Why Linux Perf Inside Docker on Raspberry Pi 5 Can Mislead Developers

A developer's write-up on Qiita documented using Linux Perf inside Docker Compose on a Raspberry Pi 5 to profile containerized workloads, a setup that appears sound but carries hidden measurement gaps. Running Linux Perf without full kernel privileges inside a container limits visibility to user-space CPU cycles, while critical system-level events such as thermal throttling, cache eviction, and scheduler decisions remain largely invisible. On a Raspberry Pi 5, these blind spots are especially significant due to its shared memory architecture and aggressive thermal management on the ARM SoC. In one documented case, container-level profiling showed a steady 15% CPU utilization, while thermal throttling was simultaneously cutting effective clock speed and causing a 25% throughput drop undetected by the profiler. The broader concern is that developers often treat measurement infrastructure as secondary, producing dashboards full of plausible-looking data that may not reflect actual system behavior.