Chameth.com - posts like apple-google-aligned-incentives, docker-automatic-nginx-proxy but not debugging-beyond-the-debugger, just-a-nod, why-you-should-be-using-https

An interesting Tailscale + Docker gotcha

2026-01-29T00:00:00Z

As I’ve written about before, I use Tailscale for a lot of things. I thought I had it set up in a reasonably secure manner, but I recently noticed a problem.

I use Tailscale’s ACLs to limit what each node can access, based on the tags I apply to it. So an app node can’t access anything via Tailscale, while an integration or server node can access things tagged with either app or integration. This is expressed pretty simply in the Tailscale ACL JSON:

{
	"tagOwners": {
		// Servers that can be SSH'd into
		"tag:server": [],
		// Applications that are exposed on tailscale but never connect out
		"tag:app": [],
		// Things which talk to other services over tailscale (connecting to apps etc)
		"tag:integration": [],
	},
	"grants": [
		{
			// Users can access everything
			"src": ["autogroup:member"],
			"dst": ["*"],
			"ip":  ["*"],
		},
		{
			// Servers and integrations can access integrations and apps
			"src": ["tag:server", "tag:integration"],
			"dst": ["tag:integration", "tag:app"],
			"ip":  ["*"],
		},
	],
}

Over the past week I’ve been setting up a private Forgejo instance behind Tailscale, complete with an actions runner that runs things using a docker-in-docker container. I didn’t want the runner knowing anything about Tailscale, so I had it configured to speak to Forgejo direct over HTTP (forgejo:3000) instead of using full Tailscale HTTPS URL that I use when accessing it (http://git.example-net.ts.net/)¹.

Everything was going fine, until I forgot to do that translation… and it worked. My Forgejo action runners could access anything on my tailnet. I run several things on Tailscale that just have authentication turned off, on the basis that only authorised devices can access them. Things like the admin interface for this website. I definitely didn’t intend for any workflow I run on my git server to have access to edit my website!

This wasn’t just limited to Forgejo, either. Any docker container I was running could access the tailnet. In hindsight it’s fairly obvious why: the host is running Tailscale, connected as a node tagged with my server tag. That creates a tailscale0 interface, and automagically sets up iptables rules to route Tailscale traffic over the interface. Docker also automagically sets up iptables rules to bridge traffic, and apparently these two sets of rules interact in such a way that traffic from Docker containers is allowed to route via the tailscale0 interface.

I say it’s fairly obvious in hindsight — there’s no reason why Docker would special case any particular host interface after all — but it still feels pretty surprising. Because both bits of software inject their own iptables rules, I never really had a good mental model for how they interact. The host Tailscale node was a completely separate building block to Docker. It would be a pain to use either of them if they didn’t do these rules, but it’s also one of the reasons I don’t really like “magical” things².

There are lots of ways to fix this, but none of them feel particularly great. You can configure both Tailscale and Docker to not automatically fiddle with iptables and handle the rules yourself, but I really hate dealing with iptables³. As a stopgap I did hold my nose and add some iptables rules to drop traffic to the tailscale0 interface if it originated from the IP ranges that Docker was configured to use:

iptables -I DOCKER-USER -s 192.168.0.0/16 -o tailscale0 -j DROP
iptables -I DOCKER-USER -s 172.17.0.0/12 -o tailscale0 -j DROP

The DOCKER-USER chain is a nice little escape hatch; it comes before the main auto-generated DOCKER chain, and Docker leaves the rules in it alone.

So I added these rules, and felt pretty good about myself, and then… everything started breaking in weird ways. After some debugging I realised the problem was DNS⁴. The server’s DNS resolver is 100.100.100.100, a special Tailscale address. This is so that it can resolve tailnet hostnames via MagicDNS (and so I don’t have to configure my custom DNS servers manually on each device; Tailscale does it for me). My new iptables rules inadvertently dropped all the DNS packets coming from docker containers. D’oh.

Obviously the solution here is to double down and add MORE iptables rules:

iptables -I DOCKER-USER -s 192.168.0.0/16 -d 100.100.100.100 -o tailscale0 -j ACCEPT
iptables -I DOCKER-USER -s 172.17.0.0/12 -d 100.100.100.100 -o tailscale0 -j ACCEPT

This got DNS working again, but it all feels a bit gross. Not least because to actually make these persist you need to use iptables-save, and then you also get all the rules that both Docker and Tailscale have inserted. I came across a script to filter out the Docker ones, but… yuck⁵.

I realised a cleaner idea might just be to lock down what server tagged nodes could access within Tailscale itself. That way I can avoid touching iptables at all. It’s handy to be able to curl services from the host when debugging, but it’s not really necessary. So I removed that access, and… stuff broke again. I’m using my Forgejo instance as a registry for some of the docker images I run, so the docker daemon needs to be able to reach it. I ended up making a new tag for infrastructure, which can be accessed from server devices. This does still allow all the docker containers to reach Forgejo, but I already have it set up with appropriate access controls and public/private repository splits. Forgejo is a service designed to run publicly, so this seems a reasonable trade-off for convenience. I used the tests feature of Tailscale’s ACL config to make sure I’d got the rules right:

{
	"tests": [
		{
			// Servers can only access infrastructure
			"src":   "tag:server",
			"proto": "tcp",
			"allow": [
				"tag:infrastructure:8080",
			],
			"deny": [
				"100.84.16.43:8080",
				"me@example.com:8080",
				"tag:server:8080",
				"tag:app:8080",
				"tag:integration:8080",
			],
		},
	],
}

The whole situation still feels a bit messy. If I ever get around to switching to nftables I might loop back and manually craft some rules for routing traffic, instead of leaving Tailscale and Docker to do their own thing.

I discovered later on that logging into the container registry at forgejo:3000 actually issued a redirect to https://git.example-net.ts.net/ so this was all basically for naught… ↩︎
I didn’t use Tailscale’s MagicDNS for a long time just because the word “magic” put me off. Only when I eventually got around to learning how it worked, and seeing that it wasn’t really that magical under the hood, did I change my mind. ↩︎
nftables seems far better in lots of ways, but I can’t really be bothered migrating. Maybe next time I reimage the server for whatever reason… ↩︎
It’s always DNS… ↩︎
Again, nftables would almost certainly help here. It actually has (gasp) configuration files. But again, I really didn’t want to spend the time migrating. ↩︎

How tech companies failed to build the Star Trek computer

2025-07-16T00:00:00Z

A computer core room on the Enterprise-D

In most Star Trek series, the ship or station computer is ever-present in the background, waiting to be called on by the main characters¹. It nearly always does exactly the right thing, and there’s little limit to the functions it can perform. Take this mundane example from DS9:

KIRA: Computer, establish link with the Bajoran Medical Index for the Northwestern District.
COMPUTER: Link established.
KIRA: Access all information on Doctor Surmak Ren.
COMPUTER: There are no records matching that name.
KIRA: Try the Northeastern District, same search.
COMPUTER: Doctor Surmak Ren, currently serving as Chief Administrator of the Ilvian Medical Complex.
KIRA: Computer, open a channel to the Ilvian Medical Complex. Administrator’s office.

The computer is doing some kind of networking to a database only identified by name. It does a search and summarises the lack of results. It then repeats the process with another database, and succinctly announces the results. Finally, it opens a communication channel to a specific room in a facility, based only on its name.

This whole interaction is remarkably boring². Kira doesn’t have to know any URLs or API endpoints, or what protocol she wants to use. She doesn’t have to open a specific app and then login and then try the query again. She just says what she wants and the computer does it.

It seems like this should be one of the most easily obtainable bits of sci-fi wizardry with our current technology. We have multiple massive companies throwing lots of money at digital assistants, LLMs that are improving at an insane rate, but we’re somehow not even close to the usability or usefulness of the Trek computers. What gives?

Boring is, well, boring.

Larry Page once said something that might help explain it:

The Star Trek computer doesn’t seem that interesting. They ask it random questions, it thinks for a while. I think we can do better than that.

This is the same Larry Page that founded Google, whose mission statement is “to organize the world’s information and make it universally accessible and useful”. Of all people, surely he should find an omnipresent computer that can answer ‘random questions’ interesting?! It seems like it should be the epitome of Google’s mission!

Google’s “better than that” seems to have been to stuff LLMs into every product they can, even when you don’t want them there. Even when they’re worse than the normal content they displace. These things look exciting when they’re part of a scripted demo at Google I/O, but they fall flat and just get in the way when they’re exposed to the reality of day-to-day use.

The Star Trek computer is the opposite: it isn’t snazzy, but it is genuinely useful. That means it’s not an attractive target for the company execs who want marketing opportunities, and it’s not appealing for engineers who need to demonstrate “impact”. But even if Google did try to make the Trek computer, there are other problems…

Assistants need to be free

A significant amount of tech companies’ business models currently revolves around trapping users in walled gardens. They want you using their ecosystem; that way they get more data from you, and you’re more likely to spend more money on their other offerings that work together. There’s barely any incentive to allow any kind of interoperability with other platforms outside carefully contracted integrations.

I remember trying to help a family member move their photos from iCloud to Google Photos. At one point they turned around and said, exasperated, “why is this so hard? Aren’t they both in the cloud?!”. It’s easy to dismiss that as someone who hasn’t quite grasped the fundamental idea that “the cloud” is just someone else’s computers, but that’s not the whole story. There’s no reason why there shouldn’t be a quick and easy transfer: both services already allow uploading and downloading, there’s just no incentive for the companies involved to make it so³.

These kinds of misaligned incentives and walled garden business models cause even more problems when it comes to digital assistants. Siri is basically never going to be able to interact with, say, your Google Drive; ~~Bard~~ Gemini is never going to be able to send a message via iMessage. Even when there are appropriately blessed interactions, they’re so clunky. Can you imagine Captain Picard saying “Computer, ask the turbolift skill to take me to deck 5”?

Someone else’s computer

Software issues aside, there’s still a key difference between the Star Trek computers and our current batch of digital assistants: where they run. The Trek computers are all housed within the ship or station they serve; they can connect elsewhere to gather information, but they run entirely independently. If they go wrong, a local engineer can go in and fix things. While some of our assistants may have physical hardware in your home, they don’t work without a vast cloud apparatus behind them. If your Internet connection fails, they become paperweights. If the company running them decide to remove some functionality you depend on, you have no recourse.

That kind of helplessness isn’t limited to assistants, either. There’s a rapidly growing trend of being unable to modify or repair hardware you fully own and control. Part of this is just that they’re becoming more complex: it’s a lot harder to replace a microchip than a gear, but companies are also going out of their way to make it more difficult for users through draconian DRM regimes⁴ and aggressive intellectual property enforcement. If the US Navy can’t repair their own equipment because a corporation says so, what hope do consumers have?

We’re approaching a point where you don’t actually own anything. Software is cloud and subscription based, hardware is unrepairable. Even cars can be remotely updated and have features added or removed. The Federation wouldn’t allow a third party control over their ships⁵, so why are we so happy to put up with it in everything we consume?

A small ray of hope?

The most promising way of tackling all of these problems is through legislation. The EU’s Digital Market Act is an attempt to force ‘gatekeepers’ like Google, Apple and Meta, to allow third-party access to their services. It seems like a pretty reasonable approach, but the tech companies are unsurprisingly resisting it. Apple in particular have gone out of their way to refuse to comply, and when forced to do so have limited the functionality to people in Europe. Still, the DMA is a promising start, and if similar legislation is introduced (and robustly enforced) elsewhere it might start forcing companies to behave a bit better.

There are also smaller companies that actually do the right thing. Framework make laptops that are user-serviceable; Fairphone do the same for mobile phones. Smaller software companies provide useful, open APIs. The average person on the street will probably have never heard of these, unfortunately, but they do still exist. Maybe as the bigger tech companies tighten the screws more, people will turn to alternatives like this? Or maybe we’ll just keep accepting that our computers work for everyone but us?

Unless, of course, the computer is playing the role of the episode’s MacGuffin and has contracted space-computer-COVID or something, then it’s a lot less in-the-background. ↩︎
It’s almost like it only exists to move the plot along. ↩︎
You can generally export your data, thanks to a combination of legislation and efforts like Google’s “Data Liberation Front”, but I’ve never seen an export format that could then just be imported into an equivalent commercial product. ↩︎
Oh, you’ve changed the screen on your iPhone? Better hope it can do the secret handshake with the Apple hardware. ↩︎
I think there might actually have been an episode where that did in fact happen. We’ll just ignore that as a plot contrivance. ↩︎

Docker reverse proxying, redux

2022-12-06T00:00:00Z

Six years ago, I described my system for configuring a reverse proxy for docker containers. It involved six containers including a key-value store and a webserver. Nothing in that system has persisted to this day. Don’t get me wrong – it worked – but there were a lot of rough edges and areas for improvement.

Microservices and their limitations

My goal was to follow the UNIX philosophy of “do one thing and do it well”. Unfortunately, that doesn’t really work when applied to network services that have to interact with one and other. UNIX tools are built upon a common file system and simple data passed over STDIN. Microservices don’t have that shared foundation. You could make one: companies that use microservices in anger often have a team that deals with the “developer experience” of creating and using microservices. But as a solo developer that’s not something I wanted to spend my time doing.

This became very apparent when trying to debug issues. In the UNIX world, if your series of commands piped together does something strange you can simply echo the data at various stages. Not so much when that data is flying around on a network, going into and out of things like etcd. Trying to figure out why a certificate hadn’t been acquired was a case of searching through logs from four containers, none of which had particularly good logging. There are many ways to get insight into what’s happening with microservices but, again, that’s not something I wanted to do myself.

Over time, and with experience in other projects, I came to realise that microservices only really make sense if you’re unable to deploy a monolith. For tech companies this naturally happens when different teams contribute to the same service: splitting it out into smaller services that are wholly owned by individual teams makes sense. For solo developers, that never happens. You can still gain the other benefits of microservices – such as code separation and having clearly defined APIs – by sticking to certain coding standards.

Proxy inconveniences

As well as being unhappy with the microservice nature of the solution, I wasn’t pleased with nginx. If you requested an unknown domain, nginx would use the first server block in its config to serve a response, instead of sending an “unrecognised name” alert as I wanted. It was a minor issue, but it irked me.

So from nginx I switched to haproxy. It has a strict-sni option when configuring TLS connections which makes it behave properly. It also performs a lot better for this type of workload than nginx. All was well for a while, but then I started getting alerts that requests were occasionally failing. I couldn’t reproduce the issue, but my nightly jobs to build and push containers managed to hit it nearly every night, causing them to fail.

After some investigation, I found that the haproxy developers had refactored the header parsing code, and neglected to properly reset flags when multiple requests were sent over the same connection. There was a patch, but it wasn’t released. No problem, I thought, I’ll just cherry-pick it onto the last release… Except that haproxy use Git in the most convoluted manner I’ve ever seen – they have one repository per release. This makes it harder to patch, but it also made me question whether I trusted them to ship stable software: there were no tests for the header parsing code (which is both fundamental and finicky, the perfect target for tests), the source code management was weird, and they didn’t seem in any rush to patch this bug.

Not long after that issue, Greg managed to encounter another bug where haproxy returned a 500 error whenever the upstream server replied with a particular, perfectly valid, header. The die was cast – it was time to move to something else.

Not Invented Here syndrome

Looking for a new solution, there were many more options than back in 2016. I’m still convinced, however, that anything exposed to the Internet should not have access to run docker containers. It’s the modern equivalent of running a CGI script as root. That single requirement eliminates most off-the-shelf solutions. What do you do when nothing quite meets your specific requirements? You make something yourself! My new solution has two components: Dotege and Centauri.

Dotege is a replacement for the microservices that monitored containers and obtained certificates. It’s fundamentally a templating engine - whenever the containers change, it evaluates a template and saves the result to disk. The template has access to details about the containers, their labels, ports, and so on. Dotege can also obtain certificates from Let’s Encrypt, and raise a signal against another process whenever the template or certificates change. I used this to generate the configuration and certificates used by haproxy for a while, and more recently changed the template so that it works for Centauri.

Centauri is my own reverse proxy. It’s configured using a simple text file and can also obtain certificates from an ACME provider. It doesn’t serve static content, has no knowledge about docker, and avoids the other bells and whistles that adorn most reverse proxies. It also has good test coverage to ensure that I don’t, say, accidentally break header parsing.

As a software engineer I enjoy writing software, but I also enjoy running simple, easy to understand software. That’s what I’ve achieved here: it’s very easy to identify where the problem is if anything goes wrong, both are small Go programs rather than vast sprawling C monstrosities, and their interaction is primarily through a file written to disk that can be inspected or edited as needed.

Apple, Google and aligned incentives

2020-10-17T00:00:00Z

The HTC Dream, the first phone released running Android.

For the past decade I’ve exclusively used Android phones. I got the HTC Dream (aka the T-Mobile G1) shortly after it came out, and dutifully upgraded every 1-2 years. In that timespan I used Android as the basis for my Master’s Thesis, took a job on the Android team at Google, and eventually became a contractor specialising in Android app development. So when I switched to using an iPhone earlier this year a few people were surprised¹.

The good old days

When Android was announced in 2007 – alongside the formation of the Open Handset Alliance – it was positioned as a bastion of openness: it would be built on open standards and the operating system would be open source. At the time iPhones were strongly coupled to iTunes and Apple was exercising strict control over what app developers could do.

When the HTC Dream was released it lived up to expectations. You could write apps for it without shelling out for a Mac! You could get root access, and it was running Linux under the hood! A whole ecosystem of custom firmwares and bootloaders started to appear, thanks to the open source nature of the OS. It shipped with some Google apps, but they were just normal apps that served as examples of what could be done.

After the Dream came a line of Nexus devices. These were Android’s flagship devices, designed to show off what a good Android phone should look like. Both the hardware and software releases tended to feature interesting, useful upgrades. Some of the original open source apps were replaced with closed source, Google proprietary ones, but that was OK - the open source versions lived on in the open source project as examples of what you could do. The devices allowed flashing custom firmware, and the OS source was always released… eventually.

The downfall

The Pixel and Pixel XL

Over time, Android has got less and less free. The Nexus line of phones gave way to the Pixel range, which got rid of the clearly demarcated border where Android ended and Google began. The Android 11 highlights list is dotted with sections that say “On Pixel devices…”, and every single one is a software feature that could be implemented on any device, but Google have decided to keep it proprietary instead of releasing it as part of the open source platform.

At the same time more and more functionality has been added to Google Play Services. This was originally a shared location for Google specific services - in 2012 it merely handled some Google+ functionality and dealing with OAuth for Google accounts. These days it contains a huge swathe of Google services, as well as platform functionality such as push notifications, barcode scanning, geolocation, and so on. Google Play Services isn’t part of the Android Open Source Project, and is only available under license from Google. You can’t really have a device without these functions, so as a manufacturer you have the choice between agreeing to whatever terms Google requires² or spending an awful lot of development time creating an alternative.

While these are fairly abstract arguments about what a free platform should look like, during the same period there has been a marked restriction in how you can actually use Android devices – both as an end-user and as a developer. iOS has always had very stark restrictions on what apps can do in the background; Android started its life allowing pretty much anything, like any good general purpose computing device. This inevitably lead to lots of apps doing lots of things that ranged from stupid to mildly suboptimal, creating a tragedy of the commons amongst apps. The victim was the phone’s battery life, and Google’s solution was a progressive series of restrictions on what apps can do and when. This includes limits on when push notifications are delivered to a device, how frequently apps can wake up, and so forth. Some of these can be bypassed by the end-user, but not all, and the process is fairly cumbersome.

Aligned incentives

From my point of view, Android and iOS are now pretty much in a similar place. They’re not general purpose computers, but app consoles: much like games consoles they consist of hardware and software that the user lacks control over but accepts in order to access the library of apps/games.

Given there’s no clear winner between them in terms of hardware and software, my decision came down to a more holistic question: how aligned are their incentives to my own? Apple is a product company: they make their money by producing shiny things that people want to purchase; Google is an advertising company: they make their money by using my personal information to show me targeted adverts.

For Google, Android was originally a strategic move to ensure that Apple couldn’t dominate the mobile web, and by extension the revenue from ads. As a company, there’s nothing to really push them forward in any particular direction other than one that facilitates advertising. Of course, Google employs tonnes of good people who want to do good things which counterbalances this, but I’d still prefer to deal with a company that has an intrinsic motivation to do things I want them to do, rather than one forced to by regulation, custom, or the good intent of their employees.

So the decision became obvious: if I have no strong opinions about the software and hardware, Apple is the clear winner because their incentives are a lot better aligned to mine.

Image credits

Photo	Creator	Licence	Source
HTC Dream	Akela NDE	CC BY-SA 3.0	Wikimedia
Pixel and Pixel XL	Maurizio Pesce from Milan, Italia	CC BY 2.0	Wikimedia

Or, at least, politely feigned surprise. ↩︎
And Google’s terms were particularly onerous: as well as requiring Chrome and Google Search to be preinstalled, they prevented manufacturers from selling any devices powered by alternative versions of Android - e.g., Samsung wouldn’t be allowed to sell a refrigerator that ran Amazon’s FireOS. The European Union handed Google a $5,000,000,000 fine for this anti-competitive behaviour, which Google are still in the process of contesting. ↩︎

An introduction to containers

2020-03-01T00:00:00Z

So. Many. Containers.

I’m a huge fan of (software) containers. Most people I know fall in to one of two camps: either they also use, and are fans of, containers, or they haven’t yet really figured them out and view them as some kind of voodoo that they don’t really want or need.

I’m writing this short guide to explain a little how containers work - and how running something in a container isn’t really that much different to running it normally - to hopefully enable more people in that second group to give them a try. It’s aimed at people who have a fairly good grasp of how Linux works.

Containers are often mentioned in the same breath as VMs, which is not a helpful comparison or analogy. Think of containers as standard units of software, much like how Intermodal containers are standard units of freight transport across the world. When a company internationally ships goods in volume there isn’t a question about how they’re packaged - they go in an intermodal container. The same container can be deployed on a freight train, a lorry, or a ship. The haulage company doesn’t need to care what’s in the container because they’re completely standardised. Likewise, with software containers you don’t really need to care about what’s inside: the software you’re deploying could be written in Go, Python2, Python3, Bash, PHP, LOLCODE, or anything¹.

What does a running container look like?

When you run a container, you’re just running a process. In a lot of ways it’s not very different to what happens if you were to start the same process on the host computer.

For example I have a container that just runs cat(1). With no arguments, cat will read from stdin until it receives an EOF, so it’s handy to test with. If I run ps a on my computer, I can see the cat process in amongst everything else I’m currently running:

    PID TTY      STAT   TIME COMMAND
   7199 pts/1    Ss     0:01 /usr/bin/zsh
 323806 pts/0    Ss+    0:00 /bin/cat
 324120 pts/4    R+     0:00 ps a

The /bin/cat process is in a container, and the ps a underneath it is just running like normal on my desktop. They look very similar, right? If I look under /proc/323806 I can see all the usual attributes, the same as any other process running.

If I run ps in a container, though, it’s a different story:

    PID TTY      STAT   TIME COMMAND
      1 pts/0    Rs+    0:00 ps a

So inside the container it looks like there’s only one process running. It can’t see anything running “outside” on my desktop. The secret here is that this isn’t a special container trick: this is just a feature of the Linux kernel called namespacing.

If we go back to procfs and look at the ns/pid node we can see the process in the container is in a separate PID (process ID) namespace to the one on my desktop:

# readlink /proc/323806/ns/pid  
pid:[4026534564]
# readlink /proc/7199/ns/pid
pid:[4026531836]

Almost all² processes running ’normally’ on my desktop have the same PID namespace, whereas each container gets their own by default. PID namespaces are hierarchical: a new process is assigned a PID in its own namespace, and the parent namespace, and the grandparent namespace, and so on. That’s why I can see the process running in the container from my normal shell - the container’s namespace is a child of the main namespace all of my desktop software is running in.

Linux supports - and container software makes use of - a bunch of other namespaces too: mount points, network, UTS³, cgroups, and more. These all play a part in isolating a container from the system it is running on.

You can manually run a process with unshare(1) to “unshare” some namespaces from the parent process. For example if I run unshare -fp --mount-proc ps a, it looks very similar to running ps instead the container:

    PID TTY      STAT   TIME COMMAND
      1 pts/4    R+     0:00 ps a

So: a process running inside a container is just a heavily namespaced process running otherwise normally in the operating system. No voodoo magic here!

What about the filesystem? What are ‘images’?

Containers run in their own mount namespace meaning mount points can be different inside the container to those on the host. This means the container can have a different / mounted to the host, effectively giving it its own filesystem.

The root filesystem of the container is defined in the container’s image. If I use Docker to run a container using the Ubuntu image (docker run ubuntu), the root filesystem inside that container will resemble a minimal ubuntu install. Note that this is just the filesystem: the container doesn’t have its own kernel.

You might be thinking that sounds pretty inefficient. Downloading Ubuntu is definitely not instant, and doing it for every application you run would be insane! Quite. Containers solve this by using filesystem layers. These are stacked on top of one another to create the final filesystem. Each layer can be retrieved and cached independently of all others.

Say (for simplicity) that the Ubuntu image is a single layer. If I run one container with that image, then the layer will be downloaded and cached once. If I run three hundred containers with that image then the layer will be downloaded and cached once. Even better, if I use another image that’s based on Ubuntu but adds some software on top, only the “on top” layer will be downloaded if I already have the relevant “Ubuntu” layer cached.

If all the layers are cached, what happens when you change a file? This is dealt with using the copy-on-write technique: when you modify a file it is copied from the source layer and the changes are only made in a new layer. This is handled by the OverlayFS filesystem which is part of the mainline kernel.

When a container is running, changes made to its filesystem are temporary, and do not persist across container restarts. To persist data - or introduce new data to a container - you can mount volumes. How this works varies depending on how you’re running your container, but at the basic level it is pretty much the same as bind-mounting (mount -o bind).

You may be familiar with using chroots to change the apparent root directory of processes, perhaps with full-blown “jails” built on top. Containers offer much better isolation thanks to the use of namespaces. Instead of being constrained to a portion of the host’s filesystem, they don’t even have it mounted! Containers also get to specify their environment - if they expect in certain places, for example - instead of the sysadmin having to manually set up the chroot. Finally, containers offer much more fine-grained control over what processes can do (if you want it), and allow much more advanced use-cases such as inter-container networking.

Images and filesystems employ a little magic to ensure that layers are reusable and cacheable, but again there’s nothing terribly special about them: a container has a filesystem that appears to it to work the same way as a filesystem on the host, and it’s using a standard filesystem shipped with the kernel.

How about networking?

Again, networking is namespaced, so a container has its own network stack, its own virtual network interface, its own IP address and so on. How that network interacts with your real network depends on how you’re running the container. Docker, for example, can add iptables rules to NAT traffic between containers' networks and the outside world.

Containers can generally be connected into networks, and can communicate amongst themselves without the traffic actually leaving the host machine. This allows you to, for example, run a SQL database and connect it to a web application without ever exposing the database to the outside world. Moreover, as well as being isolated from the outside world, it’s isolated from other containers in other networks. If one of your applications has a crazy bug or is compromised, this significantly limits the damage it can do.

You have to explicitly opt in to “publishing” ports from a container, which exposes them to the outside world (either directly, or via a load balancer or some other middle-man, depending on how you’re running the container). This means you can pick and chose how the outside world sees the app you’re deploying: if it’s a web service that listens on both port 443 and port 80, you can chose to only expose the encrypted port.

If you run some containers and create some networks, you can see the interfaces and bridges on the host:

$ ip l
...
6: br-2405a8cc0445: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP mode DEFAULT group default 
    link/ether 02:42:3e:fa:23:62 brd ff:ff:ff:ff:ff:ff
8: veth8ed0735@if7: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue master br-2405a8cc0445 state UP mode DEFAULT group default 
    link/ether b2:c1:5d:55:26:9b brd ff:ff:ff:ff:ff:ff link-netnsid 2
10: veth541d84b@if9: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue master br-9d7bc4024c1a state UP mode DEFAULT group default 
    link/ether 86:aa:5f:ee:da:1a brd ff:ff:ff:ff:ff:ff link-netnsid 1

This shows a bridge and two virtual NICs, just the same as if you’d manually created them (albeit with seemingly random names). So: as before, nothing special here.

Docker? Compose? K8s? Floccinaucinihilipilification?

(OK, Floccinaucinihilipilification isn’t actually a container technology, as far as I’m aware.)

Docker is the most popular container engine - that is, the bit of software that actually runs containers. It’s responsible for setting up all those namespaces we found, downloading and caching image layers, and actually starting and stopping the processes. Docker runs as a system-wide daemon - when you run a command like docker run ubuntu it actually just instructs the daemon to do the work.

There are several alternatives to Docker for running containers; one interesting one is Podman which runs containers without a daemon. Container engines have all standardised around the same image format looked after by the Open Container Initiative, so an image you build in Docker can be used in Podman, or pretty much any other engine.

Docker compose is a tool for defining and running multiple-container applications. I mentioned earlier running a database alongside a webapp - in practice to do this you’re going to have to configure a network for them, configure a mount point for the database to persist its data on, pass credentials in to both the database and the application, and so on. Doing all that by hand is tedious and error prone.

Docker compose lets you write “compose files”, which are simple yaml descriptions of the containers you wish to run, their properties, and details about any volumes or networks you may want. Out of the box, docker-compose will create a default network for each compose file you run so the containers within it can communicate.

Kubernetes is a container orchestrator, designed to automate deployment and management of large numbers of containers. It works with Docker under the hood, but provides a huge amount of tooling on top to allow you to deploy applications and manage their dependencies. It runs across multiple physical (or virtual) machines (while still allowing containers to communicate privately), and can support massive workloads by scaling out services (running multiple copies of a container on different hosts) and load balancing. Kubernetes is sometimes shorted to k8s (as in K - 8 elided letters - s) because computer people don’t like long words.

OK, they make sense now. But why bother?

Hopefully if you’ve read this far you’ve already picked up on some of the potential benefits, but this is my personal list:

Isolation. If I run some software in a container, there is very little it can do to upset me. It’s not going to leave bits of itself all over my filesystem, it can’t steal all of the secrets in my home directory, I can even limit its CPU and memory resources if I want. If I decide to stop running it, I just delete the container and it is completely gone: no trace remains.

Ease of use. If you give me a container image I have a very good idea of how to run it already. I might need to do some minor configuration to expose ports or mount volumes, but there’s no question about how to run it, how to make it automatically start, and there’s no “installation” procedure. If I want to then swap it with an alternative (say, move from MySQL to MariaDB), it’s potentially just a case of changing the name of the image I pull.

Dependencies included. Containers just run. Python 2 software includes Python 2 and just work. Python 3 software includes Python 3 and just work. I don’t have a massive headache trying to run both at the same time, because they take care of their own messes. Similarly I’m not going to have to install npm or cargo or composer to pull in dependencies for an application: that’s going to have been done in the build process.

Reproducibility. As a fallout from having dependencies included and being isolated from everything else, containers give you amazing reproducibility. If it “works on your machine” in a container, it’ll almost certainly work in production because it’s the exact same environment.

Standardisation. At the start of this article I called containers standard units of software. One of my favourite advantages of containers is that you basically get an API to list all the software you’re running. Most container engines let you supply labels attached to containers as well, so you can add your own annotations. I use this to annotate services which expose HTTP endpoints, and I have a tool that automatically generates SSL certificates for them and configures haproxy to route traffic to them. I can’t imagine how I’d do this without containers - I imagine it’d involve a lot of manual work.

Hopefully this has helped demystify containers a little. If you feel like I’ve missed something important out, or I’ve left you more confused than when you started, feel free to drop me a note using the feedback form below.

OK, maybe you should care if you’re deploying something written in crazy languages like PHP. ↩︎
Some multi-process apps, such as web browsers, are starting to use namespaces to enhance security, as do certain package systems like Flatpak ↩︎
“Unix timesharing system”; in practice this means having a separate hostname ↩︎