February 16, 2026
Spaceship vs grocery cart energy
Docker Swarm vs. Kubernetes in 2026
One engineer ran it for $166 a year — and the comments are on fire
TLDR: One engineer says Docker Swarm ran his whole business for $166 a year and calls Kubernetes needless complexity for most teams. Comments explode: some want a Swarm revival, others insist Kubernetes wins on features, automation, and hiring—while a middle camp pushes lightweight options like k3s as a practical compromise.
A decade-long love letter to Docker Swarm just dunked on Kubernetes, and the internet showed up with popcorn. The author claims he’s run 24 containers across two continents for just $166 a year with “zero crashes,” calling Kubernetes a “spaceship to the grocery store” used by the 1% — while the other 99% pay a complexity tax and let 87% of their CPU sit idle. Cue chaos. Some cheer the underdog, with one commenter begging for a Swarm comeback and praising its simple, turn-it-on-and-go vibe.
But the pushback is loud. Critics say the take is “just wrong,” arguing Kubernetes isn’t overkill when you need team access controls, automation, and multi-tenant setups (think multiple teams safely sharing the same system). Others point out real-world hiring: Swarm “doesn’t have the mindshare,” while K8s skills pay the bills. One fan touts k3s — a lightweight Kubernetes — as the sweet spot, and another swoons over Git-powered deployments (“git push, get a website”), name-dropping FluxCD.
Meanwhile, memes fly: the VHS-vs-Beta analogy gets dusted off, “paying for knobs you’ll never turn” becomes a catchphrase, and 87% idle CPU is roasted as “Sleep Mode-as-a-Service.” The author flashes a DIY autoscaler and a 27-line file vs 170+ lines for K8s, and the crowd splits between “simplicity wins” and “power matters.”
Key Points
- •The author reports 10 years of running Docker Swarm in production with multi-node clusters, rolling updates, and automatic rollbacks, claiming zero crashes and data loss.
- •Current infrastructure is 24 containers across two $83/year VPS instances on two continents, with ~0.3% CPU usage and disaster recovery under 10 minutes.
- •The article argues Kubernetes’ advanced features benefit a small minority, while most teams incur unnecessary complexity and cost.
- •Autoscaling is presented as Kubernetes’ only meaningful feature gap versus Swarm; the article supplies a Swarm autoscaler script and deems it smarter than Kubernetes HPA.
- •Configuration comparison shows an example of roughly 27 lines in Swarm vs 170+ lines in Kubernetes, paired with a cost breakdown highlighting significant savings.