Go Runtime Profiling¶

pprof is Go's built-in profiling toolkit for collecting and analyzing runtime performance data.

Profiling endpoints are intended for development and debugging purposes only

Be aware that memory and CPU profiling can significantly impact performance and may expose sensitive information about the application.

Configure the profiling server with these environment variables:

Key	Type	Description	Default
`PPROF_ENABLED`	boolean	Enables Go's built-in runtime profiling endpoints.	`false`
`PPROF_PORT`	number	Loopback port used by the profiling server when enabled.	`6060`

Profiling server is not exposed

The profiling server is intentionally bound to loopback and cannot be exposed through Docker port publishing.

The doco-cd image does not include a shell, so collect profiles through a helper container that shares its network namespace:

docker run --rm --network container:doco-cd curlimages/curl:latest \
  -sSf 'http://127.0.0.1:6060/debug/pprof/heap?gc=1' > heap.pb.gz

Use gc=1 for baseline and memory-growth comparisons. It forces garbage collection before the profile is collected, excluding objects that are no longer reachable but have not yet been reclaimed. Omit it only when investigating short-lived allocation spikes or garbage-collection behavior.

Compare two heap profiles with the Go toolchain:

go tool pprof -base before.pb.gz heap.pb.gz

See Go's pprof documentation for all available profiles and query parameters.