-
Story
-
Resolution: Unresolved
-
Normal
-
None
-
None
-
5
-
False
-
None
-
False
-
-
To be able to more easily analyze how eBPF resources are being used during our Perf runs, we want to try adding pprof to our Perf runs: https://github.com/netobserv/netobserv-ebpf-agent/blob/main/docs/profiling.md
At a high-level, we would need to do the following:
- expose the 6060 port in flowcollector for eBPF
- have jenkins portforward an eBPF agent pod present on an infra node (due to high load we see on those nodes)
- download the appropriate profiles during the workload run
- save that profile data as jenkins artifacts so it can be downloaded and analyzed