Substantially increased memory overhead

[dreamerlzl]

Parca-agent pods seem to use more memory with v0.32.0

A memory profiling is like the following:

The total memory usage of the parca-agent pod can surpass 500MB on my side from time to time without much conf change.

[dreamerlzl]

Testing v0.34.0 for this to see whether it has been fixed...

[brancz]

Thank you for reporting! Which version were you on before, and how much was it before?

[dreamerlzl]

Thank you for reporting! Which version were you on before, and how much was it before?

I am using v0.31 before, and the memory overhead of that is within 200 ~ 350 MB

[dreamerlzl]

/debug/pprof/heap?debug=1
heap_debug_1.txt

/debug/pprof/heap
heap.gz

the alloc_space looks like this:

it seems like we can only set a higher memory limit; there are no filtering at all in opentelemetry-ebpf-profiler and all the processes' ELFs are handled and thus memory/heapIdle spikes could be high