Docker Enterprise - The Container Platform That Actually Worked (Until Docker Sold It)

Look, Docker CE was fine for fucking around on your laptop, but the moment you tried to run it in production with any kind of scale, you'd learn real quick why enterprise software costs money. Docker Enterprise wasn't Docker being greedy - it was their admission that running containers in production is harder than their marketing made it look.

The dirty secret was that Docker CE would shit itself the moment you had more than a handful of containers, needed any kind of security compliance, or god forbid, had to run Windows containers. Try explaining to your CISO why your container platform has zero access controls and you'll understand why Docker Enterprise cost $1,125 per node per year.

The Three Components That Actually Worked

Docker Enterprise had three parts that solved real problems:

Docker Engine - Enterprise was the container runtime that didn't randomly break on kernel updates. Unlike Docker CE, which would mysteriously fail with ECONNREFUSED errors after Ubuntu patches, this version actually got tested before release. I learned this the hard way when Docker CE 19.03.8 broke our entire staging environment after a routine Ubuntu kernel update - turns out the systemd integration was fucked and containers wouldn't start with some obscure Failed to create endpoint error. Docker Enterprise never had that issue because they actually tested with production kernels.

Universal Control Plane (UCP) gave you a web interface that didn't look like it was built in 2003. More importantly, it had RBAC that actually worked - you could give developers access to their namespaces without them accidentally nuking production. The GUI was slow as molasses (we're talking 15-20 seconds to load the services page), but at least it existed. And when someone inevitably tried to docker rm -f every container on the cluster, UCP would actually stop them.

Docker Trusted Registry (DTR) was basically Harbor that didn't suck. Built-in vulnerability scanning meant you'd know your base images were compromised before they hit production, not after. The image signing feature actually worked, unlike most enterprise security theater. DTR would catch shit like using Ubuntu 16.04 base images with 47 critical CVEs and refuse to let you deploy until you fixed your Dockerfile.

The Day Docker Gave Up on Enterprise

November 2019: Mirantis bought Docker Enterprise for an undisclosed amount (read: way less than Docker hoped). Docker Inc. basically admitted they couldn't figure out how to make money from enterprise customers who actually needed shit to work.

Docker kept around 750 enterprise customers - essentially everyone who'd learned the hard way that Docker CE + wishful thinking isn't a production strategy. Those customers were paying serious money ($5K-$50K+ annually) because Docker Enterprise was the difference between "containers work on my machine" and "containers work when the CFO is breathing down your neck about uptime."

Mirantis got the customers, the codebase, and most importantly, the support engineers who knew how to debug Docker networking issues at 3am. Docker got to focus on Docker Desktop and pretend that enterprise infrastructure was someone else's problem.

When Mirantis bought Docker Enterprise, everyone expected the usual acquisition nightmare: broken promises, jacked-up prices, and a slow death spiral. Instead, something weird happened - Mirantis actually made it better. Don't get me wrong, you're still paying enterprise prices for what should be basic functionality, but at least you're getting what you pay for now.

Mirantis Container Runtime: Docker That Doesn't Suck

MCR is essentially Docker Engine Enterprise with the bugs fixed and the support team that actually gives a shit. Same commands, same Dockerfiles, same everything - except when something breaks, you don't have to wait 3 days for Docker support to suggest turning it off and on again. MCR 23.0.14 has a known memory leak in the containerd integration that kills nodes after about 72 hours of uptime, but at least Mirantis acknowledges it exists instead of pretending everything is fine.

The FIPS 140-2 validation still works, which is crucial if you're dealing with government contracts or paranoid security teams. CVE patching actually happens on a reasonable timeline now - Docker used to take months to fix shit that Mirantis fixes in weeks.

Windows containers are still a nightmare, but they're a working nightmare now. Docker CE on Windows is like playing Russian roulette with your deployment pipeline. MCR on Windows actually boots consistently, which apparently counts as a miracle in the Windows container world. The MKE 3.7 → 3.8 upgrade breaks if you have custom RBAC policies with wildcards in the resource names - learned that during a 3am emergency upgrade when half our users suddenly couldn't access anything.

Mirantis Kubernetes Engine: UCP That Doesn't Hate You

MKE is what Universal Control Plane should have been from the start. The web interface doesn't take 30 seconds to load a simple page, Kubernetes integration doesn't randomly shit itself, and the RBAC actually works the way you'd expect.

They kept Swarm support because killing it would have triggered a customer revolt, but let's be honest - Swarm is dead. If you're still running Swarm in production, you're basically running on borrowed time. Mirantis knows it, you know it, everyone knows it. The migration tools to Kubernetes are solid though.

Mirantis Secure Registry (the DTR replacement) actually works better than the original. Vulnerability scanning doesn't take forever, and the UI doesn't look like it was designed in 2005. Small victories.

The Real Talk on Customer Experience

Here's the dirty secret: Mirantis hit $100M+ run rate within two years because they actually listened to customers instead of treating them like walking ATMs.

The support tiers make sense now:

• LabCare: For when your dev team needs hand-holding ($cheaper but still slow)
• ProdCare: 24x7 support that actually answers within 2-4 hours ($expensive but worth it)
• OpsCare: They basically run it for you ($very expensive but you sleep at night)

Docker's support was famously useless - you'd open a ticket about containers crashing and get back a response asking if you'd tried using Kubernetes. Mirantis support actually understands their own product, which is apparently revolutionary in enterprise software. MKE's web interface randomly logs you out during upgrades though, so have the CLI ready.

The roadmap is honest: Kubernetes first, Swarm maintenance mode, Windows containers "we'll try our best." No bullshit about revolutionary new paradigms or synergistic cloud-native transformations.