Docker Cloud

» see cloud.docker.com as it lived, on the wayback machine

~ the last words ~

On 20 March 2018 Docker posted that the cluster and application management services in Docker Cloud would stop working on 21 May. The announcement said sixty days. The calendar said sixty-two. Either way it was two months to move whatever you had running, and the thread that formed on Hacker News that afternoon was mostly people counting production services out loud.

The product itself was straightforward and, for a while, the obvious answer to a real question. You had a Dockerfile and an image in Docker Hub and no idea what to do next. Docker Cloud connected to your AWS or DigitalOcean account, provisioned nodes into it, deployed your containers onto them, watched the health checks and scaled the count when you asked. First node free, two cents an hour for the rest.

Not all of it went at once. The registry storage and the automated builds carried on and were eventually pulled into Docker Hub, which is where the surviving parts of the brand still live. What died on 21 May was the half that made Docker Cloud a platform rather than a place to keep images.

~ the record ~

~ what tutum was ~

Before it had Docker's name on it, this was Tutum, a startup out of New York and Madrid with eleven people and a container-management service that already worked. Docker bought the company on 21 October 2015. Borja Burgos-Galindo, one of the founders, announced the rebuilt version in March 2016 as an expansion of Tutum with tighter Docker Hub integration.

Buying an operating layer is the standard move for a company that owns a format. Docker owned the image and the daemon and had no product that took money from the people using them, and Tutum was the shortest route to one. That reasoning was sound in 2015 and had eighteen months of life in it.

~ the sixty days ~

The complaints were about the window, not the decision. Docker's suggested path was to export your stacks as Compose files and run them yourself on Docker Swarm, which meant taking on the node management the product had existed to remove. Third parties moved fast: Jelastic, Cloud 66 and a run of smaller hosts published migration guides inside a fortnight, most of them written as much for acquisition as for charity.

The context nobody put in the announcement was that Docker had already lost the argument. Kubernetes took the orchestration layer over 2017 and by early 2018 the major clouds each shipped a managed version of it. A proprietary Docker-only cluster manager with a per-node price was competing against something the hyperscalers were prepared to run at cost.

~ where i was standing ~

Containers are part of how I work. The hosted layer above them never was, because the machines I care about sit on a bench and fly, and there is no region to deploy them into. Docker Cloud was aimed at people whose entire infrastructure was somebody else's computer, which in 2016 still felt like a choice rather than the default.

~ the funny part ~

cloud.docker.com still resolves. It 301s to hub.docker.com, so the address of the dead product now hands you the surviving one, which is about the politest ending a shutdown gets.

Docker's own year got worse from there. Eighteen months after Docker Cloud closed the company sold its enterprise business to Mirantis and shrank back to the registry and the developer tooling, which is roughly the shape it had before it bought Tutum in the first place.

@}-->--

              )
             (_)
            .-'-.
            |   |
            |   |
            |   |
            |   |
          __|   |__   .-.
       .-'  |   |  `-:   :
      :     `---'     :-'
       `-._       _.-'
           '""""""

« previous  ·  random  ·  next »