How Docker broke in half

&gt; “The biggest mistake was to miss Kubernetes. We were in that collective thought bubble where internally we thought Kubernetes was way too complicated and Swarm would be much more successful,” Jérôme Petazzoni, one of Docker’s first and longest serving employees, said. “It was our collective failure to not realize that.”<p>They were not wrong on saying that Kubernetes was very complicated, at least in some sense. In the beginning no one wanted to use it because they could easily setup Docker Swarm with minimal effort. This argument still pops up frequently on HN when there&#x27;s a new post about Kubernetes.<p>I guess the problem was they didn&#x27;t realize why Kubernetes <i>needs</i> to be that complicated. And if a system is complicated for good reasons, that&#x27;s actually good business opportunity and tons of people and companies will be willing to make the effort to fill that gap.

Good read, and touches on a lot of the pain points from that era. As someone who was lurking on HN at the time during the containerization boom, I think that the key failing of dotCloud&#x2F;Docker was not capitalising on Docker Swarm almost immediately. Docker Swarm was touted almost from the start but the repeated delays gave it a reputation of smoke and mirrors left people scrambling for solutions.<p>I also clearly remember the multiple high profile spats that &#x27;shykes had on HN which burned a lot of bridges. At the time he had a reputation for answering lots of questions on HN which helped a lot with community building. After those bridges were burned there was no one to speak for Docker as the developer mindsets shifted slowly towards Kubernetes. IIRC the Github PRs were also a source of contention as dotCloud corporatised.<p>To be fair, Kubernetes was a real slog to understand at the start and had a lot of competition; it was definitely not the same level of simple, direct technical solution that Docker was.<p>Interesting trip down memory lane and what a pivotal technology! Regardless of the rest Docker is a true cultural phenomenon and a testament to the insight of the creators working outside of the myopia of big tech.

&gt; The truth is, Docker had the chance to work closely with the Kubernetes team at Google in 2014 and potentially own the entire container ecosystem in the process. “We could have had Kubernetes be a first-class Docker project under the Docker banner on GitHub. In hindsight that was a major blunder given Swarm was so late to market,” Stinemates said.<p>&gt; Craig McLuckie, Kubernetes cofounder and now vice president at VMware, says he offered to donate Kubernetes to Docker, but the two sides couldn’t come to an agreement. “There was a mutual element of hubris there, from them that we didn’t understand developer experience, but the reciprocal feeling was these young upstarts really don’t understand distributed systems management,” he told InfoWorld.<p>The article criticizes Docker Swarm as myopic, but IMO, there were only two possibilities for Docker to move forward; either they acquired Kubernetes, which was a possibility in this telling of events, or they won with their own Docker Swarm.

We used Tutum before it was acquired by Docker and turned into Docker Cloud. Whilst it probably would never have scaled to our complexity needs indefinitely as we grew, it was very clear that as soon as it was purchased the product stopped going in a positive direction.<p>The only changes that they really rolled out were replacing the older, perfectly functional UI with something superficially shinier but much, much harder to use.<p>Not only was it just hard to see what you wanted to see once they&#x27;d halved the information density and got rid of any visual hierarchy, you had absolutely crazy things like a slider for the number of instances of a service that would instantly apply with no confirmation and, even crazier, responded to the scroll wheel. We once doubled our number of production pods and didn&#x27;t notice for a few hours, because someone scrolled up the page and their cursor went over the slider. Lucky they weren&#x27;t scrolling the other way.<p>Anyway, I lost a lot of trust in Docker as an organisation that knows what people value about their products as a part of that.

I don&#x27;t like so much the idea pushed by the article author that it is partly the fault of the software to be open source!<p>At least, the founders of Docker are honest and clear in my opinion:<p>&lt;&lt; Hykes disagrees with this assessment. “I think that is wrong and generally speaking the core open source product created massive growth which created the opportunity to monetize in the first place,” he said. “Lots of companies monetize Docker successfully, just not Docker. There was plenty to monetize, just Docker failed to execute on monetizing it.&gt;&gt;<p>There is still a thing that is missing there: the article let think that docker was an innovation out of nowhere and they had an unique idea that was kind of spoiled.<p>But, one has to remember the context of the period when Docker was created:<p>It was a time were cgroups and &quot;namespaces&quot; of network, process, ... were the hot new things inside the Linux and everyone was thinking about the concept of &quot;container&quot;, one way or another.<p>There were already chroots and a lot of persons were already working on &quot;app&quot;&#x2F;&quot;module&quot;&#x2F;&quot;package&quot; systems using aufs&#x2F;overlaysfs layers.<p>So, this was the moment and a few competitor were emerging like lxc and docker. Docker was very good to take the of the media cover and hype and so to become the dominant way to have &quot;containers&quot;. But, at the beginning, there was nothing particular in the technology and lots of competitors could have done it if it was not Docker.

The real problem is that docker swarm is nowhere close to being production ready and it&#x27;s full of gotchas and design decisions that make little sense as an end user. They simply couldn&#x27;t make a product good enough for the enterprise, like k8s did. I have similar reservations towards docker; sure it was a mostly novel concept (I saw live when Solomon announced docker and relatively few people knew at the time about Linux native containers or about bsd jails) but it so full of quirks and gotchas and bad APIs that it&#x27;s not smooth sailing.<p>Unfortunately they couldn&#x27;t make docker into a product (or they would have faced the wrath of the OSS community) so they probably felt stripped for time on both docker and on swarm, which made for an unpolished experience.<p>I&#x27;ve been using swarm for a small project for funz and because I didn&#x27;t want to run the full k8s (even k0s or k3s would have been heavy for my use case) and because I had bad experiences with nomad.<p>My list of complaints: - Stacks are not working 1-1 with Compose - docker machine is a separate binary - You can&#x27;t pull the latest of a docker image or the cache hit will just fail to redeploy your changes (you&#x27;re forced to do a sort of blue-green deployment or just update a tag everytime) - You have to configure, run, garbage collect a registry from scratch - docker-flow-proxy should be included out of the box - I need a way to integrate secrets<p>Don&#x27;t get me wrong, there&#x27;s an amazing amount of progress in docker swarm (and I&#x27;m running it in production, for free, on a 5$ machine). Years ago it was even worse but it&#x27;s not something I would ever recommend to my employer. I hope we&#x27;ll get there eventually.

I am a huge fan of Docker the tool, and during the prime of Docker the company they were doing some really impressive stuff. Swarm was dramatically easier to use than Kubernetes. The people working on Docker both in and outside of the company were top notch and they should be proud of what they accomplished.<p>It&#x27;s too bad they didn&#x27;t make the business side of things work, though I agree that at the time there was a certain feeling of too much money chasing an uncertain future for an open source project. I hope this next iteration of the company works out for them.

There is no mention on LXC&#x2F;jails which is ofcourse the biggest inspiration for docker, if I am not mistaken it also used lxc under the hood. That was already a good&#x27;ish product and but hard to configure and not for the mainstream at the time. Docker introduced AUFS and downloading of images which was added. I always saw docker as a properly marketed nicely ribboned lxc but mediocre implemented since they removed a lot of options at that time that were super useful like cpu limiting etc from std lxc.

Very well written article! Personally, I&#x27;m still waiting for docker swarm. The thing is enterprise doesn&#x27;t care about developer experience. They don&#x27;t need to; they pay developers so much, they have over qualified developers who can grok Kubernetes, and can hire more at a moment&#x27;s notice. They care much more about whether it meets their goals. Smaller companies&#x2F;orgs are what care about DX. They have likely very few, potentially not super highly qualified developers, and very limited resources. To them, something that their developers can set up, understand, and maintain, that&#x27;s reliable and resilient, is a HUGE win.

As someone only casually knowledgable in this area (e.g. I use docker for some of my home automation thingies, I still haven&#x27;t wrapped my head fully around what Kubernetes <i>is</i> beyond something like a way to manage a bunch of similar containers? VMs? eg)..<p>Docker, the kabillion dollar company, <i>never had a chance.</i> We all recognize the potential and actual problems when some very crucial piece of infrastructure is only free&#x2F;open source and not maintained -- but this is what happens when it is &quot;overmaintained&quot; I.e. it makes sense to invest big in something that is good and could be crucial; but only if you can both keep up quality AND stave off competitors -- and in this case, competitors aren&#x27;t just other companies, but <i>free software in general.</i> It&#x27;s kind of funny, in a way, Docker made itself a target for &quot;good competition&quot; by being so visible, in a way that e.g. &quot;curl&quot; didn&#x27;t.<p>Either way, I <i>always</i> knew &quot;Docker the kabillion dollar company&quot; was a stupid stupid bet. Small company, great bet. But this? Destined to fail.

My issue with Docker Inc has been with the support. I started to give them money once but pulled back after it became clear the support experience was essentially “customer support thyself”. I don’t have to pay anyone to have my support ticket ignored, I can do that myself for free.<p>Anyway they are past the point of no return now. Their creations will undoubtedly outlive them - probably Queen Elizabeth also. Anything they pull back will be replaced by the community. Anything new will get aped by everyone else trying to make a name for themselves. Containers are more or less a generic commodity now.

I am surprised everyone is blaming K8s for eating Docker&#x27;s lunch but from my experience in the banking industry.. it was cloudfoundry that stole the market away from Docker swarm.I don&#x27;t think I ever saw a Docker swarm in production; IT departments standardized on cloudfoundry to manage their &quot;private&#x2F;on-prem cloud&quot;.

I have always been somewhat skeptical of Docker, for example due to how horribly it has always worked on Windows. My first real exposure to containers was when getting into Red Hat certifications and being introduced to Podman, and with the benefit of hindsight I don&#x27;t really understand the benefit of Docker today when comparing the two. Podman seems just as easy, more restricted and more in line with what some might call &quot;Unix philosophy&quot;.<p>Of course when it comes to distributed environments I&#x27;m sure Kubernetes still reigns supreme, but for relatively simple container setups I don&#x27;t think there&#x27;s a need to even run Docker anymore on Linux.<p>Obviously, I&#x27;m still not well versed in containers so this might all be overlooking something.

Use the right tools for the job. I continue to run small to medium scale deployments with a couple dozen containers with Swarm. It&#x27;s smooth, and just works. Kubernetes would work too, and I use it for newer deployments.<p>Heck, I use plain old docker-compose to manage production servers with half a dozen containers, trouble-free for a few years now. Compose v2 seems promising, I&#x27;ve been using it for a while and it&#x27;s pretty stable too.<p>Swarm could have been an industry in itself, a de-facto standard for orchestration if it had been marketed and supported right. It was a lot more stable than Kubernetes in the early days and things just worked with docker while you&#x27;d have to kludge a few things to get Kubernetes doing its thing right.

kubernetes makes simple deployments look complex, and complex deployments look simple. Docker would be my choice for a simple deployment (like server + db) Only use kubernetes when you need to do complex stuff, it does make things look simple then.

But did kubernetes really won the orchestration war or is it still an ongoing thing?

Wow this was a great read! I remember taking a kubernetes class in late 2015 and the instructor then was saying that Kubernetes had already won over Docker Swarm and Mesos.<p>If I were at the helm of Docker today, I&#x27;d focus on Enterprise customers. Provide an on-prem version of DockerHub and you&#x27;ll convince thousands of companies tired of Artifactory to switch immediately. Don&#x27;t get me wrong Artifactory is pretty nice, but it&#x27;s a bear to run at scale on an enterprise level. DockerHub already is handling that traffic, so there&#x27;s no reason to think they can&#x27;t port that to an on-prem offering.

&gt; The combination of huge amounts of venture funding, a quickly growing competitive landscape, and the looming shadow of cloud industry giants all wanting a piece of the pie created a pressure cooker environment for the young company to operate within.<p>I actually got an AWS ad on the same page while reading this comment. It&#x27;s definitely hard to compete when your competitors(with lot of capital) can easily integrate your tech into theirs in no time.

I really enjoyed reading this history of the creation of Docker and the company that supports it (Docker, Inc., previously dotCloud, Inc.)

How many employees does Docker has&#x2F;had? I feel like if I got a couple hundred million dollars in funding, I could make it last many many years, producing what Docker has produced. But of course this feeling isn&#x27;t based on any entrepeneurial experience, so I&#x27;m probably wrong. But that seems like an awful lot of money to shoot through.

ugh docker only ever did one thing right, allowed me to run postgres + mysql on my laptop without the DBs shitting on the rest of the laptop<p>if they had just released a chroot database runner, would have been as good as what they did<p>it has never been good at build or deployment. it has <i>definitely</i> never been good at cloud. even kube is still awful at most things<p>these systems have never been sure if they&#x27;re configuration languages, buildsystems, plugin hosts, or operating systems, and so they&#x27;ve been bad at all 4

I wonder if there&#x27;s more room for things like fly dot io but with things like minio&#x2F;s3 or lambda included as well for bandwidth savings, kind of an &quot;aws lite&quot;

If software development was a real engineering discipline docker wouldn&#x27;t exist. The idea that you need isolated environments to run&#x2F;develop programs that need different versions of libraries is regressive. Way back in the 80s and 90s Microsoft had backwards compatiblity down. Now in the 2010s and 2020s every couple months we can expect a breaking change in the webdev, linux, python ecosystem. Try running a 3 year old python app outside a container. This foot gunning is dumb. But it sure does make work for software developers. Who go on HN and crow about how much value they create while messing around in docker or whatever.

I&#x27;m curious - does anyone have a story on how they used Docker Swarm for something at scale (&gt;100 nodes), but later migrated to K8s successfully at the same scale?

Anyone know any good metrics or analysis about trends of Kubernetes vs other ways to run containers in the not-huge-scale dev world which is 99% of software?

Kubernetes is an absolute joke. Even after being set-up right, it rarely works as it is supposed too.

The desire to change the world was stronger than the desire to make money while doing it

Happened to a pair of my Dockers once.

&gt; Hykes does acknowledge that there were tensions between the Docker and Google teams at the time. “There was a moment when egos prevailed. A lot of smart and experienced people at Google were blindsided by the complete outsiders at Docker,” Hykes said. “We didn’t work at Google, we didn’t go to Stanford, we didn’t have a PhD in computer science.<p>I found this passage interesting. Are these tensions common?

I think what happened with Docker is that it was just too good at what it was originally intended to do and too easy to use for that.<p>Programmers hate things that are easy to use (by the way I am a programmer, and no I don&#x27;t feel that way) because if they admit they use them then they might be accused of being users. And of course no programmer will admit this. They don&#x27;t realize it because it&#x27;s actually a subconscious psychological issue.<p>And so what programmers started to do was immediately make it much more complicated and at the same time, take it completely for granted. It was so useful it was like a floor to walk on. And so people started giving it the same level of respect they give a floor.<p>So for those reasons, Docker became very uncool. But at the same time it was incredibly useful. Solution: make something just like Docker but not Docker, which hip people will be allowed to use without any shame. Make it a bit more complicated and only run on the cool expensive hardware.