Question

A Go microservice running on GKE experiences steadily increasing memory usage over several hours until pods are OOMKilled and restarted. CPU utilization remains normal and request latency is stable until the moment of the restart. You have already enabled Cloud Profiler for the service. Which Cloud Profiler profile type should you analyze to most directly identify the source of the growing memory consumption?

Accepted Answer

Heap profile to see which functions retain allocated memory that is not being released

Answer

CPU time profile to find functions consuming the most processor cycles

Answer

Wall-clock (wall time) profile to measure time spent waiting on I/O and locks

Answer

Contention profile to identify goroutines blocked on mutex contention