Normalize CPU metrics for cgroups v1

appsignal / probes-rs

Rust library to read out system stats from a machine running Unix

MIT License

15 stars 2 forks source link

Normalize CPU metrics for cgroups v1 #80

Closed tombruijn closed 9 months ago

tombruijn commented 9 months ago

When we detect the number of CPUs in a container using cgroups v1, normalize the metric to total to 100%. This makes it easier to set alerts on a predicable maximum value, rather than account for the number of CPUs per container.

This change is backward incompatible for those relying on CPU totals exceeding 100%. A container with 5 CPUs will no longer have a maximum of 500% CPU usage.

Fixes #77.

backlog-helper[bot] commented 9 months ago

This Pull Request needs more reviews. @unflxw - (More info)

This is a message from the daily scheduled checks.

New issue guide | Backlog management | Rules | Feedback

backlog-helper[bot] commented 9 months ago

This Pull Request needs more reviews. @unflxw - (More info)

This is a message from the daily scheduled checks.

New issue guide | Backlog management | Rules | Feedback