search

digitalocean / vmtop

Real-time monitoring of KVM/Qemu VMs
Apache License 2.0
53 stars 19 forks source link

should exit counts capture exit reasons? #13

Open jwdeitch opened 3 years ago

jwdeitch commented 3 years ago

particularly interested in capturing the number of times the exit reason is stated as EXCEPTION_OR_NMI ? 👼 any interest in this?

jdesfossez commented 3 years ago

Yes definitely, and it would be easy to add in the current ebpf script, the issue we have is reporting, there are 60+ exit reasons IIRC and I don't think we want one column for each (either on the console or the CSV). For the CSV I think we could maybe dump a dict-like structure in a field: {HLT: 1; NMI: 42, ...} do you have another idea ? And on the console, I just don't know :-)

jwdeitch commented 3 years ago

hmm good q - I'll try the csv dump (I agree w/ the dict-like structure). not sure either about prometheus or console :) - I will give it some thought

jdesfossez commented 3 years ago

Prometheus will be fine because the exit reason can be a label so it won't show up unless it's used, the console is my biggest concern. The other tools that present exit reasons live on the console (kvm_stat, perf kvm) usually only work with a filter for specific VMs or show the sum from all the VMs. But I am fine with starting with the CSV/prometheus only and just the total on the console.