Those two GC metrics that show thread-specific data got a "thread_id" label and an updateThreadMetrics() proc, extracted from updateSystemMetrics() so it can be called in multiple threads.
I also added a new metric - nim_gc_heap_instance_occupied_summed_bytes - which sums up all the sizes we get from dumpHeapInstances(), but it turns out to be smaller than nim_gc_mem_occupied_bytes in the main thread:
Those two GC metrics that show thread-specific data got a "thread_id" label and an
updateThreadMetrics()
proc, extracted fromupdateSystemMetrics()
so it can be called in multiple threads.I also added a new metric -
nim_gc_heap_instance_occupied_summed_bytes
- which sums up all the sizes we get fromdumpHeapInstances()
, but it turns out to be smaller thannim_gc_mem_occupied_bytes
in the main thread:Do we want to keep it?