search

parca-dev / parca

Continuous profiling for analysis of CPU and memory usage, down to the line number and throughout time. Saving infrastructure cost, improving performance, and increasing reliability.
https://parca.dev/
Apache License 2.0
3.97k stars 211 forks source link

Heap profiling visualization #30

Open Slach opened 4 years ago

Slach commented 4 years ago

I like the idea about continuous profiling I hope you can implement heap profiling visualization like a stacked time-series data like this image but more detailed to each type of allocation and maybe with a link to source code

brancz commented 4 years ago

This is already available via Prometheus metrics. What are you thinking conprof to do here?

Slach commented 4 years ago

I mean something that can visualize `go tool pprof heapdata.gz not only as "top-K text-mode" console

yes it can be implemented over Prometheus metrics

but we will need too much metrics like contprof_heap{host="XXX", allocation_type="module.Type", allocation_place="source_file.go:XX"}

brancz commented 4 years ago

Yes we don’t want to reinvent the wheel but we’d love to expose this data via a prometheus remote read endpoint so it can be queried from Prometheus.

Slach commented 4 years ago

Ah, ok I understood ;)

if you will expose this data as Prometheus metrics and get an opportunity to querying over it

I will happily make some Prometheus dashboards for it