parca0.29.1

parca continuously records and analyzes how an application's CPU, memory, and I/O usage change over time, then traces those costs to method, class, and source-line level. Its profiling data gives developers a production view of hot paths, supports comparisons such as different software versions, and helps reveal wasted resources, performance regressions, memory leaks, and brief CPU or I/O spikes that affect reliability. The Parca Team publishes it as open-source Developer Tools software for Linux.

Parca Agent uses eBPF in Linux kernel space to capture the required data with low overhead, while its zero-instrumentation approach can profile infrastructure running any language or framework. Profiles can be pushed from or pulled from targets, which Parca discovers through service discovery or static configuration. A query engine designed for profiling data accepts label-selector queries and lets teams examine how code behaves across production workloads and time periods.

The software suits developers investigating resource consumption, optimizing hot paths to reduce infrastructure costs, and diagnosing problems that are difficult to reproduce on demand. Continuous collection keeps historical profiling evidence available when a regression or momentary bottleneck appears, including after a deployment changes where CPU time is spent. It requires Linux with snapd and runs on x86_64 (amd64) systems.