jwkohnen/rusage

rusage is a command wrapper like GNU time that pushes metrics of the command to a prometheus pushgateway (WIP)

rusage: a process wrapper that reports success/failure and resource usage to a prometheus pushgateway

Package rusage provides a simple process wrapper like GNU time, that reports resource usage metrics, most importantly CPU usage and maximum resident set size, start / end time and success / failure to a prometheus pushgateway. It is meant for ephemeral or batch jobs that are otherwise hard for prometheus to scrape. In the regular use case you can use the prometheus-rusage-pusher as a prefix in your command invocation line or as a Docker entrypoint and the aspect of metrics gathering is covered so you can keep the code of your workload free of prometheus details.

State

This is WIP; currently no code is released yet.

License

License under GNU Public License v3.0

rusage's Issues

This project is broken in the general case

The intention of this project was to measure maximum allocated RAM (RSS) of subtree of processes. However due to limitations of the wait() syscall family, only the cumulative maxrss of all subprocesses is available. Which is useless, because that value does not reflect the maximum of RAM all the processes have used at any given point in time. Effectively the maxrss value is only usable if and only if there is only one subprocess ever, which is a too narrow special case and too fragile for my taste.

So I won't bother any further. I've learnt a lot on the way though, so thanks for all the fish.