You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The cpu metrics were recently split from namedprocess_namegroup_cpu_seconds_total into namedprocess_namegroup_cpu_user_seconds_total and namedprocess_namegroup_cpu_system_seconds_total.
I'm wondering if it wouldn't be nicer to expose them under the old name and add an extra dimension with mode instead (basically like the node_exporter).
Then it would be backwards compatible and easier to extend if we wanted to expose other "modes" later as well...
The text was updated successfully, but these errors were encountered:
The only other modes exposed by the kernel on a per-process basis that I can see in proc(5) are guest_time and cguest_time, which I don't really see a compelling use case for.
In terms of backwards compatibility I prefer to drop metrics rather than change their labelsets. I don't like having metrics in Prometheus with a variable number of labels.
I do agree that it would be nicer to have user/system as labels of a metric rather than distinct metrics as they are now, just so you can sum() them. Now I just have to think of a new name for it...
The cpu metrics were recently split from
namedprocess_namegroup_cpu_seconds_total
intonamedprocess_namegroup_cpu_user_seconds_total
andnamedprocess_namegroup_cpu_system_seconds_total
.I'm wondering if it wouldn't be nicer to expose them under the old name and add an extra dimension with
mode
instead (basically like the node_exporter).Then it would be backwards compatible and easier to extend if we wanted to expose other "modes" later as well...
The text was updated successfully, but these errors were encountered: