The recommended way to deploy Portus is with the official Docker
image. This image is built from the
repository. Long story short, this image downloads the official
RPM and provides an
init script which is
good for any kind of container deployment.
This Docker image has in turn a tag policy worth mentioning:
head: Portus’ master branch packaged and curated as if it was production-ready. This is convenient for people that want to be on the bleeding edge and want to test the latest features. It’s not recommended to use this tag in production since it might break every now and then.
latest: the latest stable release.
- Version-specific tags (e.g.
2.3). We recommend using these tags for production clusters.
Once you have decided to use the Docker image for your deployment, you have to be aware of the following requirements:
- Configurable values follow a naming policy described
here. This page will clarify any doubts you
may have on how to configure Portus. There are some configurable values which
are important on production environments:
PORTUS_MACHINE_FQDN_VALUE: this defines the FQDN to be used. This is important because the JWT token passed between Portus and the registry relies on this setting. The default value for this will most probably not work for you.
PORTUS_CHECK_SSL_USAGE_ENABLED: this is set to true by default, but you might want to set it to false in case you are not using SSL.
RAILS_SERVE_STATIC_FILES: if you want Portus to serve assets directly (e.g. you don’t have a load balancer in front of the deployment), you will want to set this to true.
CCONFIG_PREFIX: set this to
PORTUSjust to be sure (it shouldn’t be necessary, but some deployments have had weird bugs because of this in the past).
- Check the environment variables to be used for the database.
- You have to provide three environment variables which contain secret data:
PORTUS_SECRET_KEY_BASE: which will be used for to encrypt and sign sessions (you can read more about this here).
PORTUS_KEY_PATH: used to generate the private key for JWT requests (how Portus communicates with the Registry safely).
PORTUS_PASSWORD: the password of the special
Finally, you might want to take a look at some of the examples based on docker-compose that we have implemented here.
Kubernetes and Helm
Unless you are managing Docker containers manually, or you want to deploy everything in a single machine (in which case you might probably want to check this examples), you will use a Container Orchestrator. There are a wide variety of them, but in openSUSE and SUSE we quite are invested in Kubernetes (see the kubic project).
Moreover, to maintain Kubernetes applications the community has developed Helm. Because of this, we have been working on proper Helm charts to deploy Portus in your Kubernetes cluster. We are working on pushing these charts into the main repository, but for now you can use the charts from this repository.
If you simply want to explore Portus and play with it, using the development environment might be a good fit. A quick way to start a development version of Portus is to clone the git repo and perform the following:
$ docker-compose up
For more information on development environments, check our wiki.
You can also deploy Portus in a more traditional way: simply installing the code somewhere and setup a load balancer (we recommend the containerized though). In order to install Portus, you have two options:
After that, you will have to setup everything as any other Rails application. You have an example of an NGinx configuration here. This example relies on Puma, and you might want to use the default Puma configuration. The init file from the official Docker image might give you some ideas on the environment variables to be set before starting the whole thing.