This has been called the age of DevOps, and operating systems seem to be getting a little bit less attention than tools are. However, this doesn’t mean that there has been no innovation in operating systems. [Edit: The diversity of offerings from the plethora of distributions based on the Linux kernel is a fine example of this.] Fedora CoreOS has a specific philosophy of what an operating system should be in this age of DevOps.
Fedora CoreOS’ philosophy
Fedora CoreOS (FCOS) came from the merging of CoreOS Container Linux and Fedora Atomic Host. It is a minimal and monolithic OS focused on running containerized applications. Security being a first class citizen, FCOS provides automatic updates and comes with SELinux hardening.
For automatic updates to work well they need to be very robust. The goal being that servers running FCOS won’t break after an update. This is achieved by using different release streams (stable, testing and next). Each stream is released every 2 weeks and content is promoted from one stream to the other (next -> testing -> stable). That way updates landing in the stable stream have had the opportunity to be tested over a long period of time.
Getting Started
For this example let’s use the stable stream and a QEMU base image that we can run as a virtual machine. You can use coreos-installer to download that image.
From your (Workstation) terminal, run the following commands after updating the link to the image. [Edit: On Silverblue the container based coreos tools are the simplest method to try. Instructions can be found at https://docs.fedoraproject.org/en-US/fedora-coreos/tutorial-setup/ , in particular “Setup with Podman or Docker”.]
$ sudo dnf install coreos-installer $ coreos-installer download --image-url https://builds.coreos.fedoraproject.org/prod/streams/stable/builds/32.20200907.3.0/x86_64/fedora-coreos-32.20200907.3.0-qemu.x86_64.qcow2.xz $ xz -d fedora-coreos-32.20200907.3.0-qemu.x86_64.qcow2.xz $ ls fedora-coreos-32.20200907.3.0-qemu.x86_64.qcow2
Create a configuration
To customize a FCOS system, you need to provide a configuration file that will be used by Ignition to provision the system. You may use this file to configure things like creating a user, adding a trusted SSH key, enabling systemd services, and more.
The following configuration creates a ‘core’ user and adds an SSH key to the authorized_keys file. It is also creating a systemd service that uses podman to run a simple hello world container.
version: "1.0.0" variant: fcos passwd: users: - name: core ssh_authorized_keys: - ssh-ed25519 my_public_ssh_key_hash fcos_key systemd: units: - contents: | [Unit] Description=Run a hello world web service After=network-online.target Wants=network-online.target [Service] ExecStart=/bin/podman run --pull=always --name=hello --net=host -p 8080:8080 quay.io/cverna/hello ExecStop=/bin/podman rm -f hello [Install] WantedBy=multi-user.target enabled: true name: hello.service
After adding your SSH key in the configuration save it as config.yaml. Next use the Fedora CoreOS Config Transpiler (fcct) tool to convert this YAML configuration into a valid Ignition configuration (JSON format).
Install fcct directly from Fedora’s repositories or get the binary from GitHub.
$ sudo dnf install fcct $ fcct -output config.ign config.yaml
Install and run Fedora CoreOS
To run the image, you can use the libvirt stack. To install it on a Fedora system using the dnf package manager
$ sudo dnf install @virtualization
Now let’s create and run a Fedora CoreOS virtual machine
$ chcon --verbose unconfined_u:object_r:svirt_home_t:s0 config.ign $ virt-install --name=fcos \ --vcpus=2 \ --ram=2048 \ --import \ --network=bridge=virbr0 \ --graphics=none \ --qemu-commandline="-fw_cfg name=opt/com.coreos/config,file=${PWD}/config.ign" \ --disk=size=20,backing_store=${PWD}/fedora-coreos-32.20200907.3.0-qemu.x86_64.qcow2
Once the installation is successful, some information is displayed and a login prompt is provided.
Fedora CoreOS 32.20200907.3.0 Kernel 5.8.10-200.fc32.x86_64 on an x86_64 (ttyS0) SSH host key: SHA256:BJYN7AQZrwKZ7ZF8fWSI9YRhI++KMyeJeDVOE6rQ27U (ED25519) SSH host key: SHA256:W3wfZp7EGkLuM3z4cy1ZJSMFLntYyW1kqAqKkxyuZrE (ECDSA) SSH host key: SHA256:gb7/4Qo5aYhEjgoDZbrm8t1D0msgGYsQ0xhW5BAuZz0 (RSA) ens2: 192.168.122.237 fe80::5054:ff:fef7:1a73 Ignition: user provided config was applied Ignition: wrote ssh authorized keys file for user: core
The Ignition configuration file did not provide any password for the core user, therefore it is not possible to login directly via the console. (Though, it is possible to configure a password for users via Ignition configuration.)
Use Ctrl + ] key combination to exit the virtual machine’s console. Then check if the hello.service is running.
$ curl http://192.168.122.237:8080 Hello from Fedora CoreOS!
Using the preconfigured SSH key, you can also access the VM and inspect the services running on it.
$ ssh core@192.168.122.237 $ systemctl status hello ● hello.service - Run a hello world web service Loaded: loaded (/etc/systemd/system/hello.service; enabled; vendor preset: enabled) Active: active (running) since Wed 2020-10-28 10:10:26 UTC; 42s ago
zincati, rpm-ostree and automatic updates
The zincati service drives rpm-ostreed with automatic updates.
Check which version of Fedora CoreOS is currently running on the VM, and check if Zincati has found an update.
$ ssh core@192.168.122.237 $ rpm-ostree status State: idle Deployments: ● ostree://fedora:fedora/x86_64/coreos/stable Version: 32.20200907.3.0 (2020-09-23T08:16:31Z) Commit: b53de8b03134c5e6b683b5ea471888e9e1b193781794f01b9ed5865b57f35d57 GPGSignature: Valid signature by 97A1AE57C3A2372CCA3A4ABA6C13026D12C944D0 $ systemctl status zincati ● zincati.service - Zincati Update Agent Loaded: loaded (/usr/lib/systemd/system/zincati.service; enabled; vendor preset: enabled) Active: active (running) since Wed 2020-10-28 13:36:23 UTC; 7s ago … Oct 28 13:36:24 cosa-devsh zincati[1013]: [INFO ] initialization complete, auto-updates logic enabled Oct 28 13:36:25 cosa-devsh zincati[1013]: [INFO ] target release '32.20201004.3.0' selected, proceeding to stage it ... zincati reboot ...
After the restart, let’s remote login once more to check the new version of Fedora CoreOS.
$ ssh core@192.168.122.237 $ rpm-ostree status State: idle Deployments: ● ostree://fedora:fedora/x86_64/coreos/stable Version: 32.20201004.3.0 (2020-10-19T17:12:33Z) Commit: 64bb377ae7e6949c26cfe819f3f0bd517596d461e437f2f6e9f1f3c24376fd30 GPGSignature: Valid signature by 97A1AE57C3A2372CCA3A4ABA6C13026D12C944D0 ostree://fedora:fedora/x86_64/coreos/stable Version: 32.20200907.3.0 (2020-09-23T08:16:31Z) Commit: b53de8b03134c5e6b683b5ea471888e9e1b193781794f01b9ed5865b57f35d57 GPGSignature: Valid signature by 97A1AE57C3A2372CCA3A4ABA6C13026D12C944D0
rpm-ostree status now shows 2 versions of Fedora CoreOS, the one that came in the QEMU image, and the latest one received from the update. By having these 2 versions available, it is possible to rollback to the previous version using the rpm-ostree rollback command.
Finally, you can make sure that the hello service is still running and serving content.
$ curl http://192.168.122.237:8080 Hello from Fedora CoreOS!
More information: Fedora CoreOS updates
Deleting the Virtual Machine
To clean up afterwards, the following commands will delete the VM and associated storage.
$ virsh destroy fcos $ virsh undefine --remove-all-storage fcos
Conclusion
Fedora CoreOS provides a solid and secure operating system tailored to run applications in containers. It excels in a DevOps environment which encourages the hosts to be provisioned using declarative configuration files. Automatic updates and the ability to rollback to a previous version of the OS, bring a peace of mind during the operation of a service.
Learn more about Fedora CoreOS by following the tutorials available in the project’s documentation.
Sergey
Thank you for the article.
And where can I read about CoreOS + Kubernetes ?
Or maybe I don’t fully understand the purpose of CoreOS ?
Clément Verna
OKD (https://www.okd.io/) is using Fedora CoreOS so that might be a good place to look for more info
Sergey
Excuse me one more question.
Is the information in this article relevant ? Why can’t you use ordinary images from the site ?
https://getfedora.org/en/coreos?stream=stable.
Thank.
Clément Verna
You can and should use images from getfedora.org, this article is using an old image so that you can experience the auto-update feature.
Does that makes sense ?
husimo
Hello, using Fedora 33 when I try
i have
. Seems the proper command is
.
Then I have an error using the yaml provided :
Clément Verna
Thanks for pointing it out. It should be fixed now (fcct got a new release that changes the flags and the yaml formating was lost when adding the file in wordpress 🙂 )
Anders Jackson
I guessing now, but usually there are Short and Long switches. Long are usually marked with two “-” characters and Short with one “-” characters prefix.
Short switches can be combined, so Short switch -i and -n could be combined into -in. Which isn’t a Long switch,.
So, long argument –input have two – prefix. characters and Short prefix version is just -i.
Markup languages usually change two – to a dash character, which isn’t a minus character… So that is why it can be wrong of you just cutand paste Commands into articles like this.
I would guess that -input is interpreted as the five Short switches -i -n -p -u -t and not the Long switch –input (with two -).
Ignorante
What are the differences with Fedora Silverblue and Fedora IoT?
Sergey
As I understand it :
Silverblue – desktop
IOT – rasperberry, router
CoreOS – container launching.
Maybe I am wrong.
Dennis
Thanks! It’s very useful info for me!
Sergey
Maybe not quite in the subject, but what are the mechanisms of local update ?
For example, we rock a new image, place it somewhere locally in the network and it is already being updated from …
Recently, using flatpak, SilverBlue and now CoreOS I have difficulties with the organization of a local mirror …
Dennis K
Many users rely on Pulp (https://pulpproject.org) to create local mirrors of repositories. While Pulp does not support creating mirrors for Flatpak, SilverBlue or CoreOS, we would like to add support for all three.
Would you be interested in helping us understand what kind of workflows you are looking for?
Please file an issue at https://pulp.plan.io/issues/new/.
Thank you!
Artur
You don’t need to know download address for coreos-installer, you can use in example:
it will download the newest testing image to current directory and decompress it for you.
Dragnucs
I really like Fedora CoreOS. However, to be able to manage it using Ansible, one needs to install some python stuff in it.
So how can we install software using fcct and rpm-ostree, or precisly, how to manage it via Ansible?
Dragnucs
To answer my self, one can add a one-shot systemd unit that hooks after system install and run an rpm-ostree install command.