Some time ago I've written about my experiences using it in production https://nickjanetakis.com/blog/why-i-like-using-docker-compo.... Not just for my own projects but for $500 million dollar companies and more.
We have been using it for more than five years now. Before that, we had a legacy deployment model, and I do not remember a single major issue related to Docker Compose.
We use it for both staging and production environments. The same Docker image validated in staging is deployed to production. Never fails!
This is why nobody uses it. Cloud stuff has to be as baroque as possible.
What happens if it errored on deployment or after that? you wanna write custom (bash? :D) hooks for that? What about upgrading your 'very vertically scalable' box? What if it doesn't come up after the upgrade? your downtime is suddenly hours, oops.
The k8s denial is strong and now rivals frontend frameworks denial. Never fails to amuse.
In our case, the answer is not "hope and bash". We deploy versioned images, use health checks, monitor the result, and keep rollback simple: redeploy the previous known-good image/config. Host upgrades are also treated as maintenance events, with backups and a recovery path, not as something Compose magically solves.
But I think there is an opposite mistake too: assuming every production system should be operated like a high-scale tech company.
Many production workloads are boring, predictable, and business-critical. They do not need aggressive autoscaling, multi-node orchestration, or constant traffic-spike handling. They need reliable deploys, backups, monitoring, health checks, and a clear rollback path.
That is where Compose can be a good fit: simple operational model, understood failure modes, low moving parts.
Kubernetes becomes much more compelling when you actually need automated failover, rolling deploys, autoscaling, multi-node scheduling, and stronger deployment primitives.
Not needing Kubernetes is not necessarily denial, it is just choosing the complexity budget that matches the problem.
2-3 miniPCs, cloudflare, tailscale, and k3s can save (possibly tens of) thousands on SaaS products, and would probably scale you to a company of dozens AND host your product.
outline a simple real world system to illustrate?
Get a few Beelink SER5 or SER9's, install Nextcloud to cover the files, document editing, communications (to save on Microsoft 365). Then you can have Gitea (and gitea actions) for your source code and building (skipping github enterprise), Harbor to host and scan your containers, frappe for HR, etc. Pretty much anything you pay enterprise rates for, you can self-host a version that will get your company from 1 to 100s with minimal extra work. If it's not on https://github.com/awesome-selfhosted/awesome-selfhosted, you can probably vibe code it in a couple hours.
I just started to run a k3s cluster with an almost enterprise grade software factory and a few (light) production workloads on a single cheap minipc.
I have several VMs in clouds with similar k3s architecture as yours and am wondering if there are any benefits to installing longhorn vs sticking to logical (postgres, mimir, whateveritis) replication instead.
For up to low hundreds of users I think you're better off just with 1 vertically scalable box for all the officey / web server workloads.
You mitigate the hardware failure stuff with a vendor contract where you can get someone on-site and overnight you parts, and by keeping things super boring. Volume replication is not boring, avoid at all costs. NAS or SAN if you have to but all disks in the main box for as long as you can.
For 20 person SME maybe a 2-bay Synology or similar, for a heavier company a low end 2U with hardware support. Proxmox under the OS for reduced worry snapshots, rollback, backup etc. Proxmox is there for operational flexibility, resist the temptation to create a network of VMs, you just need 1 CT or VM with all the workload inside it.
For container workloads on 1 host Portainer works as well as k8s IMHO, it gives you the key property you want - you can IaC everything declaratively with terraform + compose over an API.
Caveat that if CI gets heavy you might need to scale that out but you can keep it stateless.
Zero downtime server upgrades are easy. You could make a new server, ensure it's working in private and then adjust DNS or your floating IP address to point to the new server when you're happy. I've done this pattern hundreds of times over the years for doing system upgrades without interruption and safely. The only requirement is your servers are stateless but that's a good pattern in general.
I agree with gear54us and upvoted their comment, but I also understand what the author of the root comment is saying.
I have also delivered systems using Docker Compose that are actually running in production. The point I want to make is that people may define “production” differently depending on the number of active users, operational requirements, and risk level.
To me, this debate feels similar to the broader monolith vs. microservices debate.
Seems reasonable to assume these are serious production environments, no?!
Lots of data, caching, web apps, background workers and lots of various API integrations. No fancy React front-end, no fancy crazy system architectures. Just a typical LAMP stack but running in Docker Compose, cranking away serving value to customers with very good uptime and a very low cloud cost relative to revenue. With that said, a managed database was involved but all of the web traffic was served by apps running through Docker Compose with a simple git push model of deployment that handled thousands of deployments over the years without much fuss.
Internally, though, they wanted to self-host a chat server, Apache airflow, Overleaf for collaborative editing of research proposals, three separate Git servers, a container registry, many other things, all with extremely strict multi-tenancy isolation requirements for storage and networking because they're handling customer data and their own customers audit them for it. That was a hell of a lot easier to do with Kubernetes than trying to figure out some giant universe of barely related technologies with vastly different APIs, having to buy specialized appliances for network and storage that probably also need their own control plane software hosted somewhere else.
But if you just look at "scale" as number of http requests a particular URL gets per some unit of time, the customer-facing sites have far greater scale. If you're trying to attribute revenue, beats me. They wouldn't sell anything without the customer-facing sites, but they wouldn't have anything to sell without the internal tooling. Solo web devs get into this tunnel vision view of ops because, to them, often the web site is the product. That's not the case for most businesses.
And, of course, they'd probably just use someone else's SaaS for tooling. But if you're in a heavily regulated space where that isn't possible and you have to self-host most of your business systems, then what?
I can't personally speak to what the limit of docker compose is, as I have only worked on the lower end of this: self hosting for personal use and for small internal services serving maybe 20 users.
Are you really going to try to get 4+ 9's of uptime for a small, one-off app? Do you really need to use a cloud distributed data store that only slows things down for no real gains in practice? Do you really think the cloud services are never down, and you're willing to spend a f*ck-ton of money to create a distributed app when historically an Access DB or VB6 app would have done the job?
I've moved applications deployed via compose pretty easily... compose down -t 30 then literally sftp the application to a backup location then to the new server which only needs the Docker Engine community stack installed.. then compose up -d ... tada! In terms of deployment, you can use github action runners if you want, or anything else... you can even do it by hand pretty easily.
There is a theory which states that if ever anyone discovers exactly what Kubernetes is for and how it works, it will instantly disappear and be replaced by something even more bizarre and inexplicable.
This part is easy, Kubernetes is for your CV. /s
https://github.com/daitangio/misterio
It works very well!
https://www.macchaffee.com/blog/2024/you-have-built-a-kubern...
But docker compose can actually be very sufficient for what many projects actually need.
Granted I am a guy pushing for compose based localdevs and such but going further you often just cannot beat the simplicity of doing update QA or other CI/CD workloads in compose based projects. I have had dozens of projects where we replaced flaky slow and maintenance heavy pipelines with just docker compose up --build --wait in the past years. How come you say health checks are still broken?
Some of my concerns with compose aren’t purely technical. It makes it easier to lean on local state like volumes, bind mounts, and large .env files. Similar mechanisms exist in kubernetes, but the additional setup tends to force a bit more thought about whether they’re actually needed or just a shortcut.
On the health check side, they exist, but compose doesn’t fully act on them, that's the part that is missing. There’s no built in remediation or orchestration behavior tied to health status, which is why things like https://github.com/willfarrell/docker-autoheal exist. It’s something that was never fully carried through in Docker itself.
Agree.
Plus there's the monitoring of the host that is always overlooked in articles. I've ended up chucking Monit on there to monitor disk usage et al, and also used it for monitoring compose too and restarting containers.
And then there's Healthchecks.io, and external uptime monitoring... the list goes on. Properly monitoring systems, even single server systems, is not simple.
While not built in k8s has at least velero and kasten. However they are only possible because of snapshots https://kubernetes.io/docs/concepts/storage/volume-snapshots... and kasten has a plugin like architecture (because of k8s ) that supports application specific backups. However I never found something like that for compose. And that is troublesome in bigger projects like sentry
Docker volumes (and bind mounts) however have the minor problem of being hard to get a consistent copy to without stopping the service. You can work around this by, e. G., having ZFS or btrfs as the underlying FS and making a snapshot there. Otherwise, your software (like PostgreSQL) might also have other online backup tooling.
In my setups it just was easier to use fixed paths (or relative to project dir) from a permissions management perspective. Backup tools did not always have to/should run as root which is helpful on machines providing multiple distinct services.
Putting borg or a similar tool in a container that is part of the compose manifest file can also help. I haven't seen this used in practice though yet.
For most DBs it's one or two paths in the container, and virtually all DBs vendors have a reference Docker Compose example somewhere showing volume config. I can't remember the last time I ever "natively" installed a DB personally!
I don't think there's a good one size fits all answer to whether hosting in Compose or RDS is right for you or a given project.
It's because I like keeping my servers stateless when possible. It makes it easier to upgrade them in a zero downtime way later.
If your web server has your DB too, then you can't do zero downtime system upgrades. For example I would never upgrade Debian 12 to 13 on a live server. Instead, I'd make a new server with 13, get it all ready to go and tested and then when I'm ready flip over DNS or a floating IP address to the new server. This pattern works because both the old and new server can be writing to a database on a different server.
With all that said, if you were ok with 1 server, then yeah I'd for sure run it in Docker Compose.
How would you upgrade the server running the database?
Using a managed database solves this problem, so there's that an option.
If you self host your DB, if the data is on block storage you can at least spin up a new instance and connect that storage device onto the new instance with a short period of downtime. This is usually a satisfactory level of downtime for an event that doesn't happen too frequently.
What I like about the above is it'll work with any database and avoids needing to even think about performing real-time or near real-time replication with multiple writers.
There's also the scary truth that there's a ton of stuff out there where compliance requirements aren't enforced. I'm not saying it's a good idea but you can choose not to upgrade too. This is a risk assessment you'd need to do. At the very least if you go down this route, please make sure your server doesn't even have a public IP address. If it's super locked down, that doesn't mean it's safe but you'll want to limit the number of attack vectors as much as you can.
It really comes down to YMMV... Sometimes for a singular app surface, it's easier to just use a compose file that includes the database. mailu/mailcow is a good example... you don't necessarily want to comingle email on the same server as other services.
That said, if you need to share a single DB or set of DBs across an applicaiton with several instances/deployments, then it makes much more sense for a central deployment. I almost never do my own host level install, instead relying on cloud hosting and mgt. The only real exception is MS-SQL on internal servers... MS-SQL in Docker is barely acceptable for dev, and missing a few key features you may actually want/need.