Best Docker Containers in 2026: 12 Images Worth Running on a VPS

The best Docker containers for a VPS depend on what you intend to put online. For a website, that might mean Nginx and PostgreSQL. For a small development team, Gitea could be the entire reason to rent the server. Installing a dozen useful applications is easy enough; finding a reason to keep all twelve running is harder.

This list covers web traffic, application data, monitoring, and a few services you can use directly. Some belong together. Others do the same job in different ways, particularly the three web proxies near the beginning.

One naming detail before the list: an image is the package you download. A container is a running instance made from that image. You can replace the container during an update, provided the files you need to keep have been stored outside its disposable writable layer.

Pick the job first

1. Nginx

Nginx is straightforward to place in front of a small website. Requests arrive on the public port; its configuration decides whether to return a static file or pass the request to an application. Two hostnames can point to the same VPS and reach different applications behind it.

For routes that change occasionally, a configuration file is convenient. You can read it, review a proposed change, and see exactly where a request should go. There is no need for the proxy to discover containers automatically if the application addresses are already known.

That is the main distinction from Traefik below. Nginx can serve far more complicated deployments, but a static website and an API don’t require a complicated routing arrangement. Certificate renewal needs to be included in your setup; simply putting Nginx in a container doesn’t arrange it.

2. Traefik

If each deployment means another edit to the proxy configuration, Traefik is worth examining. With its Docker provider, routing rules can travel with the application as container labels. Traefik discovers the services and uses those rules to send requests to them.

This changes who maintains the routes. The person defining the application deployment can also define its hostname, instead of handing every change to someone managing a separate proxy file.

The integration needs access to Docker’s API. Treat that access as a privileged connection: mounting the socket read-only does not prevent the client from issuing modifying API requests. The provider documentation describes ways to restrict exposure. Its dashboard should have restricted access too.

3. Caddy

Caddy is often appealing for a much smaller reason: you want a hostname to serve an application over HTTPS without maintaining a lengthy configuration. A short Caddyfile can describe the reverse proxy, while Caddy handles certificate issuance and renewal when the relevant challenge requirements are met.

Keep its certificate state in persistent storage. Otherwise, replacing a perfectly healthy container can throw away information the next instance needs.

Between these first three choices, routing style is a useful deciding factor. Stable, explicit configuration points toward Nginx; Docker discovery points toward Traefik. Caddy suits a compact configuration with automatic HTTPS. There is little benefit in running all three for one ordinary website.

4. PostgreSQL

A shop needs to retain orders. A project application needs to relate users to tasks, permissions, and comments. PostgreSQL handles this kind of structured data with relational tables and transactions, and many self-hosted applications already support it.

The application’s compatibility requirements should decide the initial PostgreSQL version. A newer database release is no advantage if the application hasn’t added support for it.

Upgrades are where the image/container distinction becomes particularly useful. You can replace the software package, but an existing database directory may still need a migration before a different major release can use it. The official image instructions also vary by release, including storage layout. Follow the instructions for the version on your server rather than copying an old volume example.

5. Redis

Redis often arrives as an application dependency, so there may be no separate product decision to make. What does need a decision is how much of its data you can afford to lose.

For a cache, an empty restart might mean slower requests until the cache fills again. For sessions, users might have to sign in. A queue is different: losing entries can mean jobs never run. Configure persistence and recovery around the actual role, especially when Redis holds information that cannot be rebuilt from another database.

Keep it reachable only by the services that need it. A cache rarely needs its own public-facing port.

6. Portainer CE

Portainer gives Docker a browser interface. Someone investigating a stopped container can open its logs and inspect the environment without working through a terminal, which helps when server administration is shared by people with different levels of command-line experience.

For a team already deploying everything through Compose, adding a second way to change the environment may be unnecessary. Decide where configuration changes will be recorded so the live server and the deployment files don’t gradually disagree.

Portainer is an administration tool. Access to it needs the same care as access to the Docker environment it controls.

7. Uptime Kuma

SUptime Kuma answers a small but useful question: can this endpoint be reached? It can check a website or TCP service and send notifications when a check fails.

Put it on another machine if you want it to report a VPS outage. On the same host, it goes offline with the website and cannot send the alert. An internal check can still be useful, but it observes a different route from the one your visitors use.

8. Nextcloud

Nextcloud is for the people using the server, not just the people administering it. They can sync files and share them through an application hosted on infrastructure you control.

Storage is likely to shape this deployment. A shared folder that begins with a few documents can accumulate photographs, large uploads, and retained versions, so the occupied space may exceed what users see in their current files. Quotas help make that growth manageable.

Behind the interface are a database and background jobs, with caching components where the installation uses them. Backups need a consistent set of files, configuration, and database contents. Restoring only the file directory will not recreate the complete service.

9. Gitea

The container landscape is consFor a team that needs private repositories, issues, and pull requests, Gitea can provide a relatively small Git service. The attraction is having those everyday collaboration features together without adopting a larger development platform.

Repository directories contain the code history, but they aren’t the whole installation. Users, permissions, project information, attachments, and enabled storage features also belong in the recovery plan. Git access can use HTTP or SSH; choose the methods the team will actually use and configure those endpoints.

10. Vaultwarden

SecuVaultwarden is an unofficial Bitwarden-compatible server. It can make self-hosted password management practical on a modest VPS, though running a vault creates an unusual recovery problem.

Imagine that the VPS is gone and the credentials needed to rebuild it were kept inside that vault. You need an independent, protected way to recover them. Set that up before moving the team’s passwords into the service.

TLS and supported account protections, including MFA, are part of the deployment. Somebody also has to own updates and restore testing. If that responsibility has no clear owner, a managed password service may be the more workable choice.

11. Prometheus

An uptime check tells you whether a service responds. Prometheus can help explain what happens before it stops responding: rising request latency, a growing queue, or memory usage approaching the host’s limit. The available measurements depend on the exporters and application instrumentation you configure.

Be selective about labels. Putting a unique request ID into a metric can generate a separate time series for every request, turning monitoring into a storage and memory problem. Retention settings determine how much history you keep.

12. Grafana

Grafana turns data from Prometheus and other sources into dashboards. It needs those sources; installing Grafana alone doesn’t start collecting application metrics.

A useful first dashboard might show request volume, latency, and host resource usage together. When a deployment changes response times, those charts give you somewhere to investigate. Add other panels as questions arise instead of importing a large dashboard whose measurements you haven’t configured.

Running the stack on a VPS

Once you’ve identified the perfect containers for your project, consider how they’ll interact and operate at scale. Container Start with the application and its dependencies, then add the monitoring you need. A small web service might use one proxy and PostgreSQL. Redis belongs there if the application requires it; Nextcloud doesn’t belong there merely because it is useful to somebody else.

For every service that stores data, record its persistent directories or volumes. Keep secrets out of public configuration repositories, restrict administration endpoints, and test the ports reachable from outside the VPS. Docker networking can make exposure different from what you expect from local firewall settings alone.

Updates should be deliberate. Image tags can move, including tags that look like fixed version numbers. A digest identifies a particular image, making a deployment reproducible, but someone must still update the pinned reference when you adopt a newer build.

One change from the earlier list concerns MinIO. Its public repository was archived on April 25, 2026, and states that it is no longer maintained. Caddy takes its place in this selection; it is a web server, not an object-storage replacement. Existing MinIO deployments need their own support or migration assessment.

A server for the applications you chose

Docker containers aren’t just theoretical concepts; they’re revolutionizing how businesses approach application development and With a shortlist in hand, you can estimate the memory, disk space, and traffic the deployment needs. Hostomize provides VPS hosting for running Docker and your chosen applications. Allow for growing data when selecting a plan, then adjust sizing using measurements from the live services.

Visit Hostomize