Kubernetes Events Receiver Benchmark Test : Memory Behaviour

[Radhe1002]

I've been benchmarking the Kubernetes events receiver in OpenTelemetry with a load of 100 events per second. Initially, the memory usage is around 70-80 MB when the event generator starts. However, as the generator continues, the memory usage gradually increases over the first hour, stabilizing at around 1.7 GB.

When I stop the event generator, the memory usage decreases gradually over the next hour, stabilizing at 600-700 MB. It remains stable at this level for the following 10 hours, without returning to the initial 70-80 MB when no events are being generated.

I used pprof to analyze the memory usage, and it seems that the majority of the memory is being utilized by cache, JSON processing, and reflect.

Does anyone have insights into why the memory usage doesn't return to the initial levels when there are no events being processed? Could this be to caching mechanisms or memory allocation Telemetry?

[srikanthccv]

The right place for this discussion is the upstream repo where the component is maintained https://github.com/open-telemetry/opentelemetry-collector-contrib.