Starbeamrainbowlabs

Stardust
Blog


Archive

Mailing List Articles Atom Feed Comments Atom Feed Twitter Reddit Facebook

Tag Cloud

3d account algorithms android announcement architecture archives arduino artificial intelligence artix assembly async audio bash batch blog bookmarklet booting c sharp c++ challenge chrome os code codepen coding conundrums coding conundrums evolved command line compilers compiling compression css dailyprogrammer debugging demystification distributed computing documentation downtime electronics email embedded systems encryption es6 features event experiment external first impressions future game github github gist graphics hardware hardware meetup holiday holidays html html5 html5 canvas infrastructure interfaces internet io.js jabber javascript js bin labs learning library linux low level lua maintenance manjaro network networking node.js operating systems performance photos php pixelbot portable privacy problem solving programming problems project projects prolog protocol protocols pseudo 3d python reddit reference release releases resource review rust secrets security series list server software sorting source code control statistics storage svg technical terminal textures three thing game three.js tool tutorial tutorials twitter ubuntu university update updates upgrade version control virtual reality virtualisation visual web website windows windows 10 xmpp xslt

Read / Write Disk Performance Testing in Bash

Recently I needed to quickly (and non-destructively) test the read / write performance of a flash drive of mine. Naturally, I turned my attention to my terminal. This post is me documenting what I did so that I can remember for next time :P

Firstly, to test the speed of a disk, we need some data to test with. Since lots of small files will inevitably cause slowdowns due to the overhead of writing the file metadata and inode information to the superblock, it makes the most sense to use one gigantic file rather than tons of small ones. Here's what I did to generate a 1 Gigabyte file filled with zeroes:

dd if=/dev/zero of=/tmp/testfile.bin bs=1M count=1024

Cool. Next, we need to copy it to the target disk and measure the time it took. Then, since we know the size of the file (1073741824 bytes, to be exact), we can calculate the speed at which the copy took place. Here's my first attempt:

time dd if=/tmp/testfile.bin >testfile.bin

If you run this, you might find that it doesn't take it very long at all, and you get a speed of something like ~250MiB / sec! While impressive, I seriously doubt that my flash drive has that kind of speed behind it. Typically, flash memory takes longer to write to and read from - and I'm pretty sure that it can't read from it that fast either. So what's going on?

Well, it turns out that Linux is caching the disk write operations in a buffer, and then doing them in the background for us. Whilst fine for ordinary operation, this doesn't give us an accurate representation of how fast it's actually writing to the disk. Thankfully, there's something we can do about this: Use the sync command. sync will flush all cached write operations to disk for us, giving us the actual time it took to write the 1 GiB file to disk. Here's the altered command:

sync;
time sh -c 'dd if=/tmp/testfile.bin >testfile.bin; sync'

Very cool! Now, we can just take the time it took and do some simple maths to calculate the write speed of our disk. What about the read speed though? Well, to test that, we'll first need to clear out the page cache - another one of Linux's (many) caches that holds portions of files that have recently been accessed for faster retrieval - because as before, we're not interested in the speed of the cache! Here's how to do that:

echo 1 | sudo tee /proc/sys/vm/drop_caches

With the correct cache cleared, we can test the read speed accurately. Here's how I did it:

time dd if=testfile.bin of=/dev/null

Fairly simple, right? At a later date I might figure out a way of automating this, but for the occasional use now and again this works just fine :)

Found this useful? Got a better way of doing it? Want to say hi? Post in the comments below!

Rendering LaTeX documents to PDF: Attempt #2

It was all going rather well, actually - until I discovered that pandoc doesn't support regular bibliographies / references. Upon discovering this, I ended up with a bit of problem. Thankfully, the answer lay in pdflatex - but getting to the point where I could use it without having it crash on me (which, by the way it can't accomplish properly - it gives an exit code of 0 when crashing! O.o) was not a trivial journey.

This blog post is a follow up to my first post on rendering LaTeX documents with pandoc, and is my attempt to document what I did to get it to work. To start with, I installed texlive properly. Here's how to do that on apt-based systems:

sudo apt install texlive-latex-extra --no-install-recommends

The no-install-recommends is useful here to avoid ~450MiB of useless documentation (in PDF form, apparently) being dumped to your hard drive. I've also got an arch-based system (it's actually Artix Linux, that I've blogged about) which I've done this on, so here's the install command for those kind of systems:

sudo pacman -S texlive-latexextra

Once that's installed, we can use it to render our LaTeX document to PDF. Upon discussing my issues with my Lecturer at University, I discovered that you actually have to run 3 commands in succession in order to render a single PDF. Here they are:

bibtex filename
pdflatex --output-directory=. filename.tex
pdflatex --output-directory=. filename.tex

The first one compiles the bibliography using BiBTeX. If it isn't installed already, you might need to search your distribution's repositories and install it. Next, we run the LaTeX file through pdflatex from TeXLive not once but twice - as it apparently needs to resolve the references on the first pass (why it can't do them all in once pass I have no idea :P).

It's also worth noting that the bibtex command doesn't like you to append the filename extension - it does it automatically, apparently.

That's about everything I've got on the process so far. If you've got anything else to add, please let me know in the comments below (I'm rather new to this whole LaTeX thing....)

Further Reading

Rendering LaTeX documents to PDF on Linux (and maybe Windows too)

I'm starting to write another report for University, and unlike other reports, this one apparently has to be a rather specific format. To that end, I've got two choices, apparently: Use the provided Word / LibreOffice template, or use a LaTeX template instead. After the trouble and frustration I had with LibreOffice for my previous report, I've naturally decided that using the LaTeX template might be a good idea.

After downloading it, I ended up doing some research and troubleshooting to get it to render properly to a PDF. Now that I've figured it out, I thought I'd share it here for anyone else who ends up experiencing difficulties or is unsure on how it's done.

The way I'm going to be using it is with a tool called Pandoc. First, install it like so:

sudo apt install pandoc texlive-fonts-recommended

Adjust as necessary for your distribution - Windows users will need to read the download instructions. The texlive-font-recommended package is ~66MiB(!), but it contains a bunch of fonts that are needed when you're rendering LaTeX documents, apparently.

With the dependencies installed, here's the command to convert a LaTeX document to a PDF:

pandoc -s input.tex -o output.pdf

Replace index.tex with the path to your input file, and output.pdf with the desired path to the output file. I haven't figured out how to set the font to sans-serif yet, but I'll probably make another post about it when I do.

Found this helpful? Still having issues? Let me know below! I don't have analytics on here, so that's the only way I'll know if anyone reads this :-)

Sources and Further Reading

Jump around a filesystem with a bit of bash

Your shell in the middle of a teleporter.

(Banner remixed from images found on openclipart)

I've seen things like jump, which allow you to bookmark places on your system so that you can return to them faster. The trouble is, I keep forgetting to use it. I open the terminal and realise that I need to be in a specific directory, and forget to bookmark it once I cd to it - or I forget that I bookmarked it and cd my way there anyway :P

To solve the problem, I thought I'd try implementing my own simplified system, under the name teleport, telepeek, and telepick. Obviously, we'll have to put these scripts in something like .bash_aliases as functions - otherwise it won't cd in the terminal itself. Let's start with teleport:

function teleport() {
    cd $(find . -type d | grep -iP "$1" | head -n1);
}

Not bad for a first attempt! Basically, it does a find to list all the subdirectories in the current directory, filters the results with the specified regex, and changes directory to the first result returned. Here's an example of how it's used:

~ $ teleport 'pep.*mint'
~/Documents/code/some/path/pepperminty-wiki/ $ 

We can certainly improve it though. Let's start by removing that head call:

function teleport() {
    cd $(find . -type d | grep -m1 -iP "$1");
}

What about all those Permission denied messages that pop up when you're jumping around places that you might not have permission to go everywhere? Let's suppress those too:

function teleport() {
    cd $(find . -type d 2>/dev/null | grep -m1 -iP "$1");
}

Much better. With a teleport command in hand, it might be nice to inspect the list of directories the find + grep combo finds. For that, let's invent a telepeek variant:

function telepeek() {
    find . -type d 2>/dev/null | grep -iP "$1" | less
}

Very cool. It doesn't have line numbers though, and they're useful. Let's fix that:

function telepeek() {
    find . -type d 2>/dev/null | grep -iP "$1" | less -N
}

Better, but I'd prefer them to be highlighted so that I can tell them apart from the directory paths. For that, we've got to change our approach to the problem:

function telepeek() {
    find . -type d 2>/dev/null | grep -iP "$1" | cat -n | sed 's/^[ 0-9]*[0-9]/\o033[34m&\o033[0m/' | less -R
}

By using a clever combination of cat -n to add the line numbers and a strange sed recipe (which I found in a comment on this Stack Overflow answer) to highlight the numbers themselves, we can get the result we want.

This telepeek command has given me an idea. Why not ask for an index to jump to after going to the trouble of displaying line numbers and jump to that directory? Let's cook up a telepick command!

function telepick() {
    telepeek $1;
    read -p "jump to index: " line_number;
    cd $(find . -type d 2>/dev/null | grep -iP "$1" | sed "${line_number}q;d");
}

That wasn't too hard. By using a few different commands rather like lego bricks, we can very easily create something that does what we want with minimal effort. The read -p "jump to index: " line_number bit fetches the index that the user wants to jump to, and sed comes to the rescue again to pick out the line number we're interested in with sed "${line_number}q;d".

Pepperminty Wiki CLI

The Pepperminty Wiki CLI. in a terminal window, with a peppermint overlaid in the top left of the image.

I've got a plan. Since I'm taking the Mobile Development module next semester, I'd like to create an Android app for Pepperminty Wiki that will let me edit one or more instances of Pepperminty Wiki while I'm, say, on a bus.

To this end, I'll need to make sure that Pepperminty Wiki itself is all ready to go - which primarily entail making sure that its REST API is suitably machine-friendly, so that I can pull down all the information I need in the app I build.

Testing this, however, is a bit of a challenge - since I haven't actually started the module yet. My solution, as you might have guessed by the title of this blog post, is to build a command-line interface (CLI) instead. I've been writing a few bash scripts recently, to I tried my hand at creating something that's slightly more polished. Here's a list of the features supported at the time of posting:

  • Listing all pages
  • Viewing a specific page
  • Listing all revisions of a page
  • Viewing a specific revision of a page

Support for searching is on the cards, but it's currently waiting on support for grabbing search results as json / plain text from Pepperminty Wiki itself.

I'll be updating it with other things too as I think of them, but if you'd like to give it a try now, then here's the source:

It should update dynamically as I update the script. Simply save it to a file called peppermint - and then you can run ./peppermint to get an overview of the commands it supports. To get detailed help on a specific command, simply run ./peppermint {command_name} to get additional help about that specific command - and additional help for that command will be displayed if it supports any further arguments (it will be executed directly if not).

Sound interesting? Any particular aspect of this script you'd like explaining in more detail? Want to help out? Leave a comment below!

The Great Migration of Manjaro

The artix linux logo. It's relevance will become clear by the end! :P

It was just before lunch in the library, and I was checking my university emails on my travelling laptop that runs Manjaro OpenRC. While that was going on, I was inducing a few updates that it notified me about - and I started to install them with yaourt -Syua. First mistake.

During the installation, it decided to upgrade OpenRC to the version in the AUR (Arch User Repository), but I didn't think anything of it particularly - I knew that Manjaro OpenRC was dying deprecated. Second mistake.

Once the updates were complete, I shut it down and sent on my way - or at least I tried to - it wouldn't shut down, instead proceeding to log out and leave it at that. I resolved to investigate the problem when I got home. Third mistake.

By the time I came to use it again, I was greeted with an ominous message:

[Firmware Bug]: TSC_DEADLINE disabled due to Errata; please update microcode to version: 0x52 (or later)
Failed to execute /init (error -2)
Kernel panic - not syncing: No working init found. Try passing init= option to kernel. See Linux Documentation/admin-guide/init.rst
CPU: 0 PID: 1 Comm: swapper/0 Not tainted 4.13.2-1-MANJARO #1
Hardware name: Entroware Apollo/Apollo, BIOS 1.05.05 04/27/2017

Hrm. That's odd. Maybe something went wrong in the update? Linux has what's called kernel parameters that tell it how to boot. They specify things like "here's the root partition of the system", and "please let me edit files on the system after booting". To understand how this fits into the next part of the story, it's first necessary to take a quick retour and look at how, precisely the linux kernel goes about booting a system. This is best explained with a diagram:

The linux kernel boot process.

(Rendered with Ascidia. Textual diagram source available here)

  1. BIOS / UEFI POST - The starting point of the boot process. The BIOS / UEFI turns on all the devices, runs some basic hardware checks, and (usually) gives the user a choice of what they want to boot from.
  2. rEFInd - grub may be used instead of rEFInd, but the basic principle is the same: it asks the user how they want to boot from the hard drive. Kernel parameters are decided on here.
  3. Initialisation: The Linux kernel is executed by the bootloader, and it proceeds to initialise itself and the connected devices.
  4. Mount initial RAM disk: The Linux encounters a chicken-and-egg problem rather early on: How can it start talking to the connected devices if it doesn't know how to talk to them? The initial RAM disk solves that problem: It contains a bunch of drives and other such components that the kernel needs to initialise all the connected devices. It's like a cut-down root file system, in a sense.
  5. Load drivers: The Linux kernel loads the drivers from the initial RAM disk (aka initrd) and starts initialising all the connected devices.
  6. Mount root (read-only): The main root file system is mounted next, but only in read-only mode while the boot process finishes.
  7. Execute init: It is at this point that the very first process is executed. It usually presides at /sbin/init, but this can be changed through the init kernel parameter.
  8. Mount root (read-write): The init process (under SysVinit at least) then remounts the root filesystem such that it is writeable.
  9. Mount other partitions: The next job is the mounting of the other partitions in /etc/fstab. This is also done by SysVinit if I recall correctly.
  10. Reach runlevels: The main runlevels managed by the service manager (e.g. OpenRC) are now executed in order by the service manager.

Phew, that took more explaining than I thought! And to think it all happens in the span of about 10 seconds....! With that out of the way, let's continue with the story.

Let's try specifying the init kernel parameter - maybe the update cleared it for some random reason....? I had no idea what I was getting myself into :P

Unexpectedly, specifying init=/sbin/init didn't work. Neither did specifying init=/bin/sh. At this point, I suspected that there was something seriously wrong. I (correctly) guessed that it was the update I performed that morning that was to blame. After a bunch of backing and forthing, I managed to get hold of a previous copy of the openrc package that was replaced by the 0.27 version from the AUR. After doing a full backup, I tried installing it and removing the new openrc-sysvinit package that was also installed.

Before we continue further, I should probably explain how I managed to install the previous package version. Didn't I just explain that my system wasn't bootable? Well, yes. But I also had the original manjaro-architect installation media that I used to build the system in the first place. With that in hand, I could use rEFInd to boot from that (my UEFI firmware makes it a bit of a pain otherwise!), and then mount the root partition of the broken system and chroot into it. This process allows me to pretend that the system is actually booted, while piggybacking off the live installation media of the boot process. It works a bit like this:

lsblk # Find the root partition
mkdir /mnt/os;
mount /dev/sdZY /mnt/os # Mount the root partition
mount /dev/sdAB /mnt/os/boot/efi # Mount the EFI partition
manjaro-chroot /mnt/os bash # Enter the chroot and execute bash

Back to the story. Sadly, valiant though my effort was to replace the openrc and openrc-sysvinit packages was, it did not solve the problem. Eventually, I ended up having to perform a blind migration to Artix Linux, the spiritual successor to both Manjaro OpenRC and Arch OpenRC (apparently the developers of both came together to create Artix Linux).

Eventually, I ended up with a successful migration that I performed inside the chroot, and the system was bootable again! Next time, I'll always run pacman -Syu before yaourt -Syua. I'll also set up a temporary backup solution for my system files (I've already got one in place for my personal files) while I figure out a more permanent one that backs up across the network.

Sources and Further Reading

Run a program on your dedicated AMD graphics card on Linux

I've recently figured out how to run a program on my dedicated AMD R7 M445 graphics card in Ubuntu 17.04, and since it's taken me far too long to around figuring it out, I thought I'd note it down here for future reference - if it helps you too, let me know in the comments below!

It's actually really simple. First, check that your dedicated AMD graphics card shows up with lspci:

lspci

If it's anything like my setup, you'll get a pair of rows like this (though they might not be next to each other):

00:02.0 VGA compatible controller: Intel Corporation HD Graphics 620 (rev 02)
01:00.0 Display controller: Advanced Micro Devices, Inc. [AMD/ATI] Topaz XT [Radeon R7 M260/M265 / M340/M360 / M440/M445] (rev c3)

Thankfully, my dedicated AMD card is showing (better than it did in previous versions of ubuntu, too, which thought it was an M225!). Next, we need to check that the amdgpu kernel module is loaded with a quick lsmod:

lsmod | grep -i amd

On my laptop, I get this:

amdkfd                139264  1
amd_iommu_v2           20480  1 amdkfd
amdgpu               1564672  1
i2c_algo_bit           16384  2 amdgpu,i915
ttm                    98304  1 amdgpu
drm_kms_helper        151552  2 amdgpu,i915
drm                   352256  9 amdgpu,i915,ttm,drm_kms_helper

Yay! It's loaded. Now to do a test to see if we can run anything on it:

glxinfo | grep "OpenGL renderer"
DRI_PRIME=1 glxinfo | grep "OpenGL renderer"

The above runs glxinfo twice: Once on the integrated graphics card, and once on the dedicated graphics card. The key here is the DRI_PRIME=1 environment variable - this tells the amdgpu driver that this process should run on the dedicated graphics and not the integrated graphics card. On my machine, I get this output:

OpenGL renderer string: Mesa DRI Intel(R) HD Graphics 620 (Kabylake GT2) 
OpenGL renderer string: Gallium 0.4 on AMD ICELAND (DRM 3.9.0 / 4.10.0-33-generic, LLVM 4.0.0)

As you can see, the latter invocation of the command ran on the dedicated AMD graphics card, and the former on the integrated graphics. So simple!

Now that we've verified that it works, we can do it with any program:

DRI_PRIME=1 inkscape

Did this you find this helpful? Did it work (or not)? Let me know in the comments!

Sources

Unmounting NFS Shares on Shutdown in OpenRC Manjaro

A cool SVG of a server. (Above: A clipart image of a server. Source)

Since I've been using Manjaro with OpenRC when I'm out and about, I've been steadily fixing little issues and niggles I've been encountering one by one (such as finding the option to let you move the windows on the taskbar panel around yourself).

One of the first issues I encountered was that OpenRC would generously take the network down before my NFS (network file system) shares have been unmounted. This results in lengthly delays when shutting down as each of the components of the NFS mounting system have to be waited upon by OpenRC and finally killed after taking too long to shut down.

Initially I attempted to investigate reordering the shutdown process, but that quickly grew out of hand as I was investigating, and I discovered that it was not a particularly practical or, indeed, stable solution to my particular problem. Next, I found autofs which looked like it would solve the problem by automatically mounting and unmounting my NFS shares as and when they are needed, but despite assisance from someone far more experienced in the Manjaro world than I (thank you!) couldn't get it to work reliably. In addition, it started exhibiting some odd behaviour like hiding all my other mounts in my /media folder, so I went on the hunt for better solution.

Quite by chance (all thanks to Duck Duck Go Instant Answers!) I stumbled upon NetworkManager dispatcher scripts. NetworkManager is the service / application that manages, surprisingly, the network connections on several major linux distributions - including Ubuntu (which I've used before), and, crucially, Manjaro. Although the answer said that the functionality I wanted had been removed, upon looking into the amtter it appeared to be an artifact of the way systemd shutdown the system, and so I gave it a whirl anyway just to see if it would work.

Thankfully it did end up working! To that end, I thought I'd (re)post the solution I found here for future reference, and in case it helps anyone else :-)

Assuming you already have your shares set up and working in your /etc/fstab, you can create a file in the folder /etc/NetworkManager/dispatcher.d/pre-down.d with the contents something like this:

#!/bin/sh

logger "Unmounting NFS shares gracefully before the network goes down...";

umount /media/bob/rocket-diagrams-nas;
umount /media/sean/satellite-schematics;

logger "Unmounted NFS shares successfully.";

Once done, you'll need to make it executable with a quick sudo chmod +x, and try rebooting to test it!

In theory, this could be used to do other things that need to be done before the network is taken down, like making a sekret tracking request to your web server for anti-theft purposes, or uploading a backup of your laptop's /etc directory automagically in case it comes to a sticky end.

Sources and Further Reading

Debug your systemd services with journalctl

Some nice bulbs and bits of wood.

The chances are that if you're using linux, you will probably have run into systemd. If you find yourself in the situation where you've got a systemd service that keeps dying and you don't know why (I've been there before several times!), and there's nothing helpful in /var/log, before you give up, you might want to give journalctl a try. It's systemd's way of capturing the output of a service and storing it in it's logging system (or something).

When I first found out about it, I read that apparently journalctl -xe servicename would show me the logs for any given service. It turned out that it wasn't the case (it just threw a nasty error), so I went trawling through the man pages and found the correct command-line switch. If you've got a service called rocketbooster.service, and you want to see if systemd has any logs stored for it, then you can execute this command:

journalctl --unit rocketbooster.service

...or for short

journalctl -u rocketbooster.service

It should open the logs (if there are any) in less - with the oldest logs at the top, so you might need to scroll all the way down to the bottom to see anything that's relevant to your problem (shift + G will take you to the bottom of the file).

I've found that systemd has a habit of rotating the logs too - and journalctl doesn't appear to know how to access the rotated logs, so it's best if you use this command as soon as possible after failure (suggestions on how to access these rotated logs are welcome! Post down in the comment :D).

I thought I'd document it here in case it was useful to anyone - and so I don't forget myself! :P

Access your home linux box from anywhere with SSH tunnels

An abstract tunnel that doesn't hold much relevant to the blog post :P

(Header by GDJ from openclipart.org. Source page)

....and other things! Recently, I bought a Raspberry Pi 3. Now that the rest of the components have arrived, I've got a rather nice little home server that's got a 1 terabyte WD PiDrive attached to it to provide lots of lovely shared storage, which is rather nice.

However, within a few weeks I was faced with a problem. How do I access my new box to configure it from my internship when I'm on lunch? Faced with such a challenge, I did what anyone would, and took to the internet to find a solution.

It didn't take long. A while ago I heard about these things called 'SSH tunnels', which, while not designed for a high throughput, are more than adequate for a low-intensity SSH connection that runs a few kilobytes a second in either direction. After reading this excellent answer by erik on the Unix & Linux StackExchange, I had an understanding of how SSH tunnels work, and was ready to put together a solution. You should go and read that answer if you'd like to understand SSH tunnels too - it explains it much better than I ever could :P

With that knowledge in hand, I went about planning the SSH tunnel. I already have a server a public IP address (it's hosting this website!), so I needed a reverse tunnel to allow me to access a port local to my linux box at home (called elessar - a virtual cookie for anyone who gets the reference!) from starbeamrainbowlabs.com.

Important! Ask yourself whether it's moral and ethical to set up an ssh tunnel before you think about following along with this article! If you find yourself behind a firewall or something similar, then the chances are that it's there for a good reason - and you might get into trouble if you try and circumvent it. I won't be held responsible for any loss or damages of any description caused by the reading of this post.

First job: create a limited account on starbeamrainbowlabs.com for elessar to SSH into. That's easy:

sudo useradd --system ssh-tunnel

Then, with a few quick lines in /etc/ssh/sshd_config:

Match User ssh-tunnel
    ForceCommand echo 'This account can only be used for ssh tunnelling.'

....we can prevent the ssh-tunnel user from being abused to gain shell access to the server (let me know if there are any further measures I can put in place here).

Now that I had a user account to ssh in as, I could set up a public / private keypair to authenticate with starbeamrainbowlabs.com, and cook up an SSH command for elessar that would set up the appropriate tunnel. After fiddling around a bit, I came up with this that did the job:

ssh -TN -R30582:localhost:5724 ssh-tunnel@starbeamrainbowlabs.com

Very cool. So with that command executing on elessar, I could ssh into elessar from starbeamrainbowlabs.com! In short, it sets up a tunnel that will make port 30582 on starbeamrainbowlabs.com tunnel through to port 5724 on elessar - the port on elessar that has SSH running on it, without allocating a pseudo-tty to save resources. explainshell.com can, well, explain it in more detail if you're interested.

Having an SSH command that would set up the tunnel is nice, but it's not very useful, since I have to execute it first before I can actually SSH into elessar from afar.

The solution was actually a little bit complicated. First, I wrote a simple systemd service file (systemd is what I have installed, since it's vanilla raspbian - this should be easily adaptable to other systems and setups) to start the SSH tunnel automagically on boot:

[Unit]
Description=SSH tunnel from starbeamrainbowlabs.com to local ssh server.

[Service]
Type=simple
ExecStart=/usr/bin/ssh -TN -R30582:localhost:5724 ssh-tunnel@starbeamrainbowlabs.com

[Install]
WantedBy=network-online.target

I quickly realised that there were a few flaws with this approach. Firstly, it tried to start the SSH connection before my router had connected to the internet, since my router starts faster than the box that initialises the fibre connection to my ISP. Secondly, it fails to retry when the connection dies.

The first problem can be solved relatively easily, by wrapping the ssh command in a clever bit of shell scripting:

/bin/sh -c 'until ping -c1 starbeamrainbowlabs.com &>/dev/null && sleep 5; do :; done && /usr/bin/ssh -TN -R30582:localhost:5724 ssh-tunnel@starbeamrainbowlabs.com

The above tries to ping starbeamrainbowlabs.com every 5 seconds until it succeeds, and only then does it attempt to open the SSH connection. This solves the first problem. To solve the second, we need to look at autossh. Autossh is a small tool that monitors an ssh connection in a variety of configurable ways and restarts the connection if ever dies for whatever reason. You can install it with your favourite package manager:

sudo apt install autossh

Substitute apt with whatever package manager you use on your system. With it installed, we can use a command like this:

autossh -o "UserKnownHostsFile /home/ssh-tunnel/.ssh/known_hosts" -o "IdentityFile /home/ssh-tunnel/.ssh/ssh-tunnel_ed25519" -o "PubkeyAuthentication=yes" -o "PasswordAuthentication=no" -o "ServerAliveInterval 900" -TN -R30582:localhost:5724 -p 7261 ssh-tunnel@starbeamrainbowlabs.com

to automatically start our ssh tunnel, and restart it if anything goes wrong. Note all the extra settings I had to specify here. This is because even though I had many of them specified in ~/.ssh/config for the ssh-tunnel user, because of systemd's weird environment when it starts a service, I found I had to specify everything in the command line with absolute paths (ugh).

Basically, the above tells autossh where the known_hosts file is (important for automation!), that it should only attempt public / private keypair authentication and not password authentication, that it should check the server's still there every 15 minutes, and all the other things we figured out above.

Finally, I combined the solutions I came up with for both problems, which left me with this:

[Unit]
Description=SSH tunnel from starbeamrainbowlabs.com to local ssh server.

[Service]
Type=simple
ExecStart=/bin/sh -c 'until ping -c1 starbeamrainbowlabs.com &>/dev/null && sleep 5; do :; done && /usr/bin/autossh -o "UserKnownHostsFile /home/pi/.ssh/known_hosts" -o "IdentityFile /home/pi/.ssh/ssh-tunnel_ed25519" -o "PubkeyAuthentication=yes" -o "PasswordAuthentication=no" -o "ServerAliveInterval 900" -TN -R30582:localhost:5724 -p 7261 ssh-tunnel@starbeamrainbowlabs.com'

[Install]
WantedBy=network-online.target

Here's a version that utilises the -f parameter of autossh to put the autossh into the background, which eliminates the sh parent process:

[Unit]
Description=SSH tunnel from starbeamrainbowlabs.com to local ssh server.

[Service]
Type=forking
Environment=AUTOSSH_PIDFILE=/var/run/sbrl-ssh-tunnel/ssh-tunnel.pid
PIDFile=/var/run/sbrl-ssh-tunnel/ssh-tunnel.pid
ExecStartPre=/bin/mkdir -p /var/run/sbrl-ssh-tunnel
ExecStartPre=-/bin/chown ssh-tunnel:ssh-tunnel /var/run/sbrl-ssh-tunnel
ExecStart=/bin/sh -c 'until ping -c1 starbeamrainbowlabs.com &>/dev/null && sleep 5; do :; done && /usr/bin/autossh -f -o "UserKnownHostsFile /home/pi/.ssh/known_hosts" -o "IdentityFile /home/pi/.ssh/ssh-tunnel_ed25519" -o "PubkeyAuthentication=yes" -o "PasswordAuthentication=no" -o "ServerAliveInterval 900" -TN -R30582:localhost:5724 -p 7261 ssh-tunnel@starbeamrainbowlabs.com'

[Install]
WantedBy=network-online.target

I ended up further modifying the above to set up an additional tunnel to allow elessar to send emails via the postfix email server that's running on starbeamrainbowlabs.com. Let me know if you'd be interested in a tutorial on this!

Sources and Futher Reading

Art by Mythdael