ps alternative to "who eats memory" #51
Description
Problems with ps.
- default output is not useful
➜ ~ ps
PID TTY TIME CMD
487827 pts/4 00:00:00 zsh
495775 pts/4 00:00:00 ps- useful command is
ps faux, but I can't explain what it does without https://explainshell.com/explain?cmd=ps+faux (and even with it) - in any case switches are not intuitive for troubleshooting server issues
- getting human readable memory stats is hackish
ps afu | awk 'NR>1 {$5=int($5/1024)"M";}{ print;}'
- no colors to help you digest the info
Draft of the ideal solution.
root@a36a405677a1:/var/log# ps faux
USER PID %CPU %MEM VSZ RSS TTY STAT START TIME COMMAND
root 2992289 0.0 0.0 4240 3448 pts/0 Ss 19:59 0:00 /bin/bash
root 2995029 0.0 0.0 5896 2896 pts/0 R+ 20:11 0:00 \_ ps faux
root 1 0.0 0.0 1104 40 ? Ss Jan11 0:26 /sbin/docker-init -- /usr/local/bin/my_init
root 6 0.0 0.1 20656 4792 ? S Jan11 0:00 /usr/bin/python3 -u /usr/local/bin/my_init
root 17 0.0 0.0 2524 88 ? S Jan11 0:10 \_ /usr/bin/runsvdir -P /etc/service
root 18 0.0 0.0 2372 72 ? Ss Jan11 0:00 \_ runsv supervisord
root 20 0.0 0.0 2608 68 ? S Jan11 0:00 | \_ /bin/sh ./run
root 22 0.0 0.4 34700 16832 ? S Jan11 9:53 | \_ /usr/bin/python3 /usr/bin/supervisord -c /etc/supervisor/supervisord.conf
azuraca+ 28 1.0 0.0 388716 3716 ? Sl Jan11 136:27 | \_ /usr/local/bin/icecast -c /var/azuracast/stations/radio_plato/config/ic
azuraca+ 30 0.2 0.0 388716 3160 ? Sl Jan11 30:21 | \_ /usr/local/bin/icecast -c /var/azuracast/stations/radio_plato_-_test_x_
azuraca+ 46 7.5 17.3 1748928 701272 ? SLl Jan11 987:47 | \_ /usr/local/bin/liquidsoap /var/azuracast/stations/radio_plato/config/li
azuraca+ 2800539 7.2 8.0 1565660 324600 ? SLl 06:03 61:09 | \_ /usr/local/bin/liquidsoap /var/azuracast/stations/radio_plato_-_test_x_
root 19 0.0 0.0 2372 72 ? Ss Jan11 0:00 \_ runsv cron
root 21 0.0 0.0 9920 944 ? S Jan11 0:02 \_ /usr/sbin/cron -f
The above, with tree sorted by memory usage in human readable form, stat legend or omitted, colors, vsz+rss merged (or replaced with mem+swap), tty removed (no idea how it is useful, but -v, --verbosity flag may turn it back).
The goal is to give a clear understanding which tree is using the most memory. Much like neofetch gives understanding which OS and hardware is used. For that it probably makes sense to sort tree such that last screen contains the branch and processes with most memory usage. It would also help to see the graph of most memory usage over time, with spikes or OOM events, maybe in interactive interface like visidata.