poor man’s ruptime

Historically the original ruptime¹ was using broadcast udp/513² in a network. Since it's not 1982 anymore, but 2022 today, here's a version for multiple networks with encrypted traffic and client-server architecture.

You will automatically get instant list of hosts (down or up), inventory of hardware, software overview, comparable list of benchmark results.

While it was

• rcp (remote copy)
• rexec (remote execution)
• rlogin (remote login)
• rsh (remote shell)
• rstat
• ruptime
• rwho (remote who)
• rwall (remote wall)

It is now

• ruptime (remote uptime) - the classic
• runame (remote uname and OS/release) - keep track what OS/release you run
• rsw (remote software) - what kind of package managers did sneak in
• rhw (remote hardware, inventory) - what hardware you have
• rload (remote load of CPU/MEM/GPU/GPUMEM) - usage of hardware
• rbench (remote benchmark) - comparable list of your hardware
• rboot (remote rebootable?) - safety level for a reboot
• rnet (remote network) - networking details (interface name, connection speed)
• rdisk (remote disk) - overview of local disks and their speeds
• rac (remote users' connect time) - overview of usage (see ac³)

Never heard of ruptime, what does it look like?

The output shows how long the system has been up, the number of users currently on the system, and the load averages⁴. The load average numbers give the number of jobs in the run queue averaged over 1, 5 and 15 minutes.

$ ruptime # FQDN   State Uptime    Users    Load Averages 1' 5' 15'
dolphin.ocean.net   up   15+05:57  0 users  load 0.04 0.08 0.07
fish.ocean.net      up    4+21:27  0 users  load 0.22 0.25 0.25
tuna.ocean.net      up    4+21:27  0 users  load 0.20 0.30 0.42

$ runame # FQDN              Kernel Release Architecture, OS Version Code
banana.ocean.net             Darwin 19.0.0 x86_64, MacOSX 10.15.1 19B77a
fish.ocean.net               Linux 5.15.0-17-generic x86_64, Ubuntu 22.04 jammy
lemon.ocean.net              GNU/kFreeBSD 11.4-0-amd64 x86_64, Debian unreleased sid
tuna.ocean.net               Darwin 21.1.0 arm64, macOS 12.0.1 21A559

$ rload # FQDN               CPU %  MEM %  GPU %  MEM %
whale.ocean.net               19.00   3.37  51.20  42.12

$ rsw # FQDN                 pkg number...
seahorse.ocean.net           dpkg 7243 rpm 0 pip3 393 

$ rhw # FQDN                 age        cores memory
fish.ocean.net               2008/09/08 8 31
lemon.ocean.net              2021/08/09 16 16
banana.ocean.net             2019/12/10 64 377

$ rnet # FQDN                interface linkspeed wlanquality%
fish.ocean.net               wlp3s0     144Mb/s 86
orca.ocean.net               enp4s0 	 2500Mb/s

$ rboot # FQDN               users   screen/tmux   cpu load
orca.ocean.net               users 2 screen/tmux 1 CPU 5

$ rbench # FQDN              Memory   Total CPU      Cores
orca.ocean.net               MEM 5.05 94 GB CPU 6.16 32

Command line options

-i   Install the software
-u   Upload information to the server (NOTE: this might require root permission and get restricted to the root user in the future)
-v   Print license/version and quit

No option queries the server for the information.

Why would I want this?

• it's simple⁵
• monitoring systems have no or not very useful CLI tools
• you don't want to manually keep a list of hosts
• you want to see what hosts are down
• you want to see what hosts are not idle
• you want to run something on all running hosts with parallel
• get rid of non-standard/in-house solutions that do not scale or are cumbersome in some other way

Real life examples

Get an overview of your operating systems and releases

$ runame | awk '{i[$NF]++} END {for (n in i) print i[n] " " n}' | sort -nr

Find hosts that are least used by CPU

$ rload | sort -k2n

Find hosts that have 90%+ usage of either CPU/MEM/GPU/GPUMEM

$ rload -c | grep " [9][0-9].\| [0-9][0-9][0-9]."

Update rnet output for all online hosts

# for a in `ruptime | grep -v " down " | awk '{print $1}'`; do echo $a; ssh root@$a "runame -u"; done

List all hosts sorted by network speed

$ rnet | sort -k3nr

Combined ruptime and rload output

$ join <(ruptime) <(rload) | column -t

Run something on all hosts having Ubuntu 22.04

# runame | grep jammy | awk '{print $1}' | parallel -j0 'ssh root@{} "something"'

Get total cores and memory of all your machines

$ rhw|awk '{print $3 " " $4}'|datamash -t" " sum 1-2

Average age of computers, oldest and newest (by BIOS date)

$ rhw|awk '{print $2}'|sed "s,/.*,,g"|datamash -t" " median 1 min 1 max 1

Right adjusted rhw output

$ rhw|column -t -R3,4

Your total diskspace

$ rdisk | sed "s,sd.,,g;s,nvme... ,,g;s,md.,,g;s,mmcblk.,,g" |sed "s,.*ethz.ch,,g" | awk '{for(i=t=0;i<NF;) t+=$++i; $0=t}1' |datamash sum 1

Sometimes nl or ts (from moreutils) are useful as well.

Configuration

The defaults for rwhod/ruptime is downtime after 11' (11*60 seconds)⁶ (ISDOWN), status messages are originally generated approximately every 3' (AL_INTERVAL)⁷.

SERVER=wedonthaveaprivacyproblem.com
PORT=51300
HOSTNAMECMD='hostname -f'

Create a key for the encryption with mcrypt. You will need this on server and client for symmetric encryption.

COLUMNS=160 dd if=/dev/urandom bs=1 count=60 2>/dev/null > /etc/ruptime/ruptime.key

Create a local user to run the daemon.

adduser --disabled-password --quiet --system --home /var/spool/ruptime --gecos "ruptime daemon" --group ruptime

Running the daemon.

daemon --user=ruptime:ruptime mini-inetd 51300 /usr/sbin/ruptimed

Classic Mode

If you set HOSTNAMECMD='hostname -s' you will have the same mode as original rwho/ruptime/rwhod. You can even limit the thing to your single one network with

iptables -A INPUT -p tcp --dport 51300 --match ttl --ttl-gt 1 -j REJECT

Requirements

• Client: nc xz bc cron ethtool dmidecode memtester lm-sensors datamash nvidia-smi timeout mcrypt wireless-tools acct
• Server: nc xz tcputils daemon mcrypt
• Optionals: pen trickle bkt iptables

Supported Systems

• macOS
• Linux
• FreeBSD
• Windows if someone implements uptime.exe (https://www.windowscentral.com/how-check-your-computer-uptime-windows-10#check_pc_uptime_cmd)

Starting it

• FreeBSD: rc.d

• Linux: daemon, init.d, cron @reboot, systemd

• macOS: https://medium.com/swlh/how-to-use-launchd-to-run-services-in-macos-b972ed1e352

• Windows (not sure if they still have net start, haven't seen it since NT 4)

• without systemd

# crontab -l
*/1 * * * *  /usr/bin/ruptime -u
*/3 * * * *  /usr/bin/rload -u
@reboot      /usr/bin/runame -u
@reboot      /usr/bin/rsw -u

Some metrics are not useful to have at regular intervals, nor at every boot, so collect them when needed, examples:

rboot -u
rnet -u

On first setup and hardware changes (memory upgrade, disks added):

rbench -u
rdisk -u
rhw -u

rwho

If you really wanted rwho, here's a hint

who |sed "s/ [a-z]/$(hostname -f):&/1;s,: ,:,1"

Other r commands

• rdate, https://github.com/resurrecting-open-source-projects/openrdate

Special Files

• https://manpages.debian.org/unstable/manpages/nologin.5.en.html
• https://manpages.debian.org/unstable/manpages/issue.5.en.html
• https://manpages.debian.org/unstable/manpages/motd.5.en.html
• https://manpages.debian.org/unstable/proftpd-basic/ftpusers.5.en.html
• https://manpages.debian.org/unstable/cron/crontab.1.en.html references cron.allow cron.disallow
• https://manpages.debian.org/unstable/login/login.1.en.html
• https://manpages.debian.org/unstable/manpages/services.5.en.html

Footnotes

1. https://en.wikipedia.org/wiki/Berkeley_r-commands ↩

2. https://manpages.debian.org/unstable/manpages/services.5.en.html ↩

3. https://manpages.debian.org/unstable/acct/ac.1.en.html ↩

4. https://www.brendangregg.com/blog/2017-08-08/linux-load-averages.html ↩

5. https://www.gkogan.co/blog/simple-systems/ ↩

6. https://sources.debian.org/src/netkit-rwho/0.17-14/ruptime/ruptime.c/ ↩

7. https://sources.debian.org/src/netkit-rwho/0.17-14/rwhod/rwhod.c/ ↩