Monet - Monitor Network

A small CLI to monitor network traffic to an external host.

Context

Currently, I Work From Home (WFH) and as such are on many video calls. I noticed that my network connection was not always stable and wanted to monitor it. Usually, I have a constant ping running to my favorite DNS provider. But, this requires actively monitoring the ping output to determine if a response hasn't come back. While this process works, I've found it quite distracting to:

recognize I might be lagging,
switching context to a terminal to monitor the ping, and
waiting for a second or so of no output to recognize I'm probably lagging.

So, my goal is to write a small CLI that can either wrap ping or do something similar and report back to me if a few pings have been missed. This way, I can keep my focus on my work and only be alerted if something is wrong.

Notes

Charting Libraries

ntcharts - had way too many options and was a bit overwhelming
asciigraph - simple enough to get things started

Errors

Left running a long time and saw a bunch of on-send-err: &probing.Packet{Rtt:0, IPAddr:(*net.IPAddr)(0xc0000a4de0), Addr:"2600:6c66:0:4::6:c", Nbytes:32, Seq:16777, TTL:0, ID:2383}; write udp [::]:181->[2600:6c66:0:4::6:c]:0: sendto: network is unreachable errors
~~Left running a long time and came back to graph with crazy axis values (really small IIRC, like huge negative numbers).~~
- ~~reproduce by pinging a docker container, kill container, restore container... shows up occasionally in a few seconds.~~
- ~~avg is fine, sd is way off when printing. -9223372036854.775808 is the number that is being printed.~~
- ~~number is most negative float you can have https://stackoverflow.com/a/56989911~~
- ~~Probably want to check and manage our own ping data + handle this timeout cleaner~~
- resolved with https://github.com/prometheus-community/pro-bing/pull/130
Graph disappears if all values are NANs (happens when there is a long outage)
We get a ton of recv: id: 6223; seq: 752; not found errors as we start getting packets outside our plotted window

Ideas

[i] Replace asciigraph with home rolled solution to be able to "XXX" out a column where we are expecting a response but didn't get one.
Slow down to a "reasonable" rate once the screen is filled with data.
Show a warning if we haven't seen a response or two in an expected time window.
Use a different intervals to make more human sense: 50ms, 100ms 250ms 500ms 1s
Add a screen to search/choose from a known list of hosts to monitor.
[?] Look into charm-bracelet's Tape library for testing + demo recording
Look into not using a ping library to implement the ping functionality
[i] Look into non-charm-bracelet UI library to reduce dependencies (low priority)
Include a histogram of the ping times
Show p90, p95, p99, p995, p999, p9995 latencies
Keep a "window" for ~1000 pings and compute more "recent" statistics
Fix negative standard deviation issue

Key:

[ ] - Not Started
[x] - Done
[~] - In Progress
[?] - Not Sure
[!] - Problem
[i] - Icebox