Go 1.27 adds goroutine leak profiler to runtime for production debugging

Go 1.27 adds a goroutine leak profiler that can be queried via the runtime/pprof package or the /debug/pprof/goroutineleak HTTP endpoint. The profile is named "goroutineleak" and works with the existing net/http/pprof infrastructure, so services that already expose pprof gain the new capability automatically. It reports goroutine stacks that are permanently blocked on synchronization primitives, allowing engineers to see which code paths are leaking.

The addition addresses a long-standing gap: production systems could not be inspected for goroutine leaks, whereas testing tools like goleak and synctest only catch leaks in unit tests. Leaked goroutines consume heap memory and increase GC cycles, especially when GOMEMLIMIT is active, leading to degraded latency and throughput. By surfacing these leaks in production, operators can intervene before resource exhaustion occurs.

The profiler’s detection scope is limited to goroutines blocked on channels or sync package primitives such as mutexes and wait groups. It does not flag goroutines blocked on OS operations like network reads or file I/O, nor can it differentiate between a temporary surge of blocked workers and a true leak. Consequently, low-volume or non-sync-primitive leaks may remain invisible.

Enabling the profiler incurs minimal runtime cost because it only records stacks of goroutines that meet the blocking criteria and does not require continuous sampling. However, teams must add the pprof handlers if they are not already present and allocate storage or monitoring pipelines to collect the generated profiles. The trade-off is a small increase in observability for a large gain in leak detection accuracy.

For engineers, the new profile changes the debugging workflow: instead of reproducing leaks in tests, they can now query live services and correlate leak reports with recent deployments or traffic spikes. The precise nature of the reports reduces the need for manual stack analysis, but teams still need to interpret whether a reported block is intentional or a bug. Integrating the profile into alerting systems can provide early warnings before leaks cause noticeable performance degradation.