Synopsis

$ stack install ...list of packages... > output 2>&1 &
$ ./monitor-stack > output.stats
...
$ ./compute-utilization
$ ./make-data-js
$ open report.html

Introduction

This repo contains scripts used to monitor a long running stack install job under Linux.

The script monitor-stack will periodically write statisics about the install job to the file output.stats. The format of each line is:

1467875898.43 27.8 6 explicit-exception-0.1.8 fail-4.9.0.0 farmhash-0.1.0.5 extra-1.4.7 Cabal-1.22.8.0 Cabal-ide-backend-1.23.0.0
^- unix time  ^- % idle
                   ^- number of active workers
                     ^- list of packages being worked on

Stats are written to the file approximately once every 5 seconds. It will automatically quit when it determines the stack install process is no longer running.

monitor-stack will us ps to find the stack install process automatically. However, if it fails to find the process you can specify its pid as the first argument:

monitor-stack pid... > output.stats

monitor-stack looks for the exact string stack install in the ps output.

compute-utilization will compute the average %-idle from the output.stats file.

make-data-js will create JSON data from output.stats for use with report.html.

report.html will show a graph of CPU utilization over the course of the job.

Caveats

• Only tested under Ubuntu Linux