• Yomi - Yet one more installer
  • What is Yomi
  • Overview
  • Installing and configuring salt-master
    • Other ways to install salt-master
    • Looking for the pillar
    • Enabling auto-sign
    • Salt API
  • The Yomi formula
    • Looking for the pillar in Yomi
    • Enabling auto-sign in Yomi
    • Salt API in Yomi
      • Real time monitoring in Yomi
  • Booting a new machine
    • The ISO image
    • PXE Boot
    • Finding the master node
    • Setting the minion ID
    • Adding user provided configuration
    • Container
  • Basic operations
    • Getting hardware information
    • Configuring the pillar
    • Cleaning the disks
    • Applying the yomi state
  • Pillar reference for Yomi
    • config section
    • partitions section
    • lvm section
    • raid section
    • filesystems section
    • bootloader section
    • software section
    • suseconnect section
    • salt-minion section
    • services section
    • networks section
    • users section

Yomi (yet one more installer) is a new proposal for an installer for the [open]SUSE family. It is designed as a SaltStack state, and expected to be used in situations were unattended installations for heterogeneous nodes is required, and where some bits of intelligence in the configuration file can help to customize the installation.

Being also a Salt state makes the installation process one more step during the provisioning stage, making on Yomi a good candidate for integration in any workflow were SaltStack is used.

To execute Yomi we need a modern version of Salt, as we need special features are only on the master branch of Salt. Technically we can use the last released version of Salt for salt-master, but for the minions we need the most up-to-date version. The good news is that most of the patches are currently merged in the openSUSE package of Salt.

Yomi is developed in OBS, and actually consists on two components:

• yomi-formula: contains the Salt states and modules requires to drive an installation. The source code of the project in available under the openSUSE group in GitHub.
• openSUSE-Tubleweed-Yomi: is the image that can be used too boot the new nodes, that includes the salt-minion service already configured. There are two versions of this image, one that is used as a LiveCD image and other designed to be used from a PXE Boot server.

The installation process of Yomi will require:

• Install and configure the salt-master service.
• Install the yomi-formula package.
• Prepare the pillar for the new installations.
• Boot the new systems with the ISO image or via PXE boot

Currently Yomi support the installation under x86_64 and ARM64 (aarch64) with EFI.

SaltStack can be deployed with different architectures. The recommended one will require the salt-master service.

zypper in salt-master

systemctl enable --now salt-master.service

For different ways of installation, read the official documentation. For example, for development purposes installing it inside a virtual environment can be a good idea:

python3 -mvenv venv

source venv/bin/activate

pip install --upgrade pip
pip install salt

# Create the basic layout and config files
mkdir -p venv/etc/salt/pki/{master,minion} \
      venv/etc/salt/autosign_grains \
      venv/var/cache/salt/master/file_lists/roots

cat <<EOF > venv/etc/salt/master
root_dir: $(pwd)/venv

file_roots:
  base:
    - $(pwd)/srv/salt

pillar_roots:
  base:
    - $(pwd)/srv/pillar
EOF

Salt pillar are the data that the Salt states use to decide the actions that needs to be done. For example, in the case of Yomi the typical data will be the layout of the hard disks, the software patterns that will be installed, or the users that will be created. For a complete explanation of the pillar required by Yomi, check the section Pillar in Yomi

By default Salt will search the states in /srv/salt, and the pillar in /srv/pillar, as established by file_roots and pillar_roots parameters in the default configuration file (/etc/salt/master).

To indicate a different place where to find the pillar, we can add a new snippet in the /etc/salt/master.d directory:

cat <<EOF > /etc/salt/master.d/pillar.conf
pillar_roots:
  base:
    - /srv/pillar
	- /usr/share/yomi/pillar
EOF

The yomi-formula package already contains an example of such configuration. Check section Looking for the pillar in Yomi

To simplify the discovery and key management of the minions, we can use the auto-sign feature of Salt. To do that we need to add a new file in /etc/salt/master.d.

echo "autosign_grains_dir: /etc/salt/autosign_grains" > \
     /etc/salt/master.d/autosign.conf

The Yomi ISO image available in Factory already export some UUIDs generated for each minion, so we need to list into the master all the possible valid UUIDs.

mkdir -p /etc/salt/autosign_grains

for i in $(seq 0 9); do
  echo $(uuidgen --md5 --namespace @dns --name http://opensuse.org/$i)
done > /etc/salt/autosign_grains/uuid

The yomi-formula package already contains an example of such configuration. Check section Enabling auto-sing in Yomi

The salt-master service can be accessed via a REST API, provided by an external tool that needs to be enabled.

zypper in salt-api

systemctl enable --now salt-api.service

There are different options to configure the salt-api service, but is safe to choose CherryPy as a back-end to serve the requests of Salt API.

We need to configure this service to listen to one port, for example 8000, and to associate an authorization mechanism. Read the Salt documentation about this topic for different options.

cat <<EOF > /etc/salt/master.d/salt-api.conf
rest_cherrypy:
  port: 8000
  debug: no
  disable_ssl: yes
EOF

cat <<EOF > /etc/salt/master.d/eauth.conf
external_auth:
  file:
    ^filename: /etc/salt/user-list.txt
    salt:
      - .*
      - '@wheel'
      - '@runner'
      - '@jobs'
EOF

echo "salt:linux" > /etc/salt/user-list.txt

The yomi-formula package already contains an example of such configuration. Check section Salt API in Yomi

The states and modules required by Salt to drive an installation can be installed where the salt-master resides:

zypper in yomi-formula

This package will install the states in /usr/share/salt-formulas/states, some pillar examples in /usr/share/yomi/pillar and configuration files in /usr/share/yomi.

Yomi expect from the pillar to be a normal YAML document, optionally generated by a Jinja template, as is usual in Salt.

The schema of the pillar is described in the section Pillar reference for Yomi, but the yomi-formula package provides a set of examples that can be used to deploy MicroOS installations, Kubic, LVM, RAID or simple openSUSE Tumbleweed ones.

In order to salt-master can find the pillar, we need to change the pillar_roots entry in the configuration file, or use the one provided by the package:

cp -a /usr/share/yomi/pillar.conf /etc/salt/master.d/
systemctl restart salt-master.service

The images generated by the Open Build Service that are ready to be used together with Yomi contains a list a random UUID, that can be used as a auto-sign grain in salt-master.

We can enable this feature adding the configuration file provided by the package:

cp /usr/share/yomi/autosign.conf /etc/salt/master.d/
systemctl restart salt-master.service

As described in the section Salt API, we need to enable the salt-api service in order to provide a REST API service to salt-minion.

This service is used by Yomi to monitor the installation, reading the event bus of Salt. To enable the real-time events we need to enable set events field to yes in the configuration section of the pillar.

We can enable this service easily (after installing the salt-api package and the dependencies) using the provided configuration file:

cp /usr/share/yomi/salt-api.conf /etc/salt/master.d/
systemctl restart salt-master.service

Feel free to edit /etc/salt/master.d/salt-api.conf and provide the required certificates to enable the SSL connection, an use a different authorization mechanism. The current one is based on reading the file /usr/share/yomi/user-list.txt, that is storing the password in plain text. So please, do not use this in production.

Once we check that in our config of the pillar contains this:

config:
  events: yes

We can launch the yomi-monitor tool.

export SALTAPI_URL=http://localhost:8000
export SALTAPI_EAUTH=file
export SALTAPI_USER=salt
export SALTAPI_PASS=linux

yomi-monitor -r -y

The yomi-monitor tool store in a local cache the authentication tokens generated by Salt API. This will accelerate the next connection to the service, but sometimes can cause authentication errors (for example, when the cache is in place but the salt-master get reinstalled). The option -r makes sure that this cache is removed before connection. Check the help option of the tool for more information.

As described in the previous sections, Yomi is a set of Salt states that are used to drive the installation of a new operating system. To take full control of the system where the installation will be done, you will need to boot from an external system that provides an already configured salt-minion, and a set of CLI tools required during the installation.

We can deploy all the requirements using different mechanisms. One, for example, is via PXE boot. We can build a server that will deliver the Linux kernel and an initrd with all the required software. Another alternative is to have an already live ISO image that you use to boot from the USB port.

There is an already available image that contains all the requirements in Factory. This is an image build from openSUSE Tumbleweed repositories that includes a very minimal set of tools, including the openSUSE version of salt-minion.

To use the last version of the image, together with the last version of salt-minion that includes all the patches that are under review in the SaltStack project, you can always use the version from the devel project

Note that this image is a _multibuild one, and generates two different images. One is a LiveCD ISO image, ready to be booted from USB or DVD, and the other one is a PXE Boot ready image.

The ISO image is a LiveCD that can be booted from USB or from DVD, and the last version can be always be downloaded from:

wget https://download.opensuse.org/repositories/systemsmanagement:/yomi/images/iso/openSUSE-Tumbleweed-Yomi.x86_64-livecd.iso

This image does not have a root password, so if we have physical access to the node we can become root locally. The sshd service is enabled during boot time but for security reasons the user root cannot access via SSH (PermitEmptyPasswords is not set). To gain remote access to root we need to set the kernel command line parameter ym.sshd=1 (for example, via PXE Boot).

The second image available is a OEM ramdisk that can be booted from PXE Boot.

To install the image we first need to download the file openSUSE-Tumbleweed-Yomi.x86_64-${VERSION}-pxeboot-Build${RELEASE}.${BUILD}.install.tar from the Factory, or directly from the development project.

We need to start the sftpd service or use dnsmasq to behave also as a tftp server. There is some documentation in the openSUSE wiki, and if you are using QEMU you can also check the appendix document.

mkdir -p /srv/tftpboot/pxelinux.cfg
cp /usr/share/syslinux/pxelinux.0 /srv/tftpboot

cd /srv/tftpboot
tar -xvf $IMAGE

cat <<EOF > /srv/tftpboot/pxelinux.cfg/default
default yomi
prompt   1
timeout  30

label yomi
  kernel pxeboot.kernel
  append initrd=pxeboot.initrd.xz rd.kiwi.install.pxe rd.kiwi.install.image=tftp://${SERVER}/openSUSE-Tumbleweed-Yomi.xz rd.kiwi.ramdisk ramdisk_size=1048576
EOF

This image is based on Tumbleweed, that leverage by default the predictable network interface name. If your image is based on a different one, be sure to add net.ifnames=1 at the end of the append section.

The salt-minion configuration in the Yomi image will search the salt-master system under the salt name. Is expected that the local DNS service will resolve the salt name to the correct IP address.

During boot time of the Yomi image we can change the address where is expected to find the master node. To do that we can enter under the GRUB menu the entry ym.master=my_master_address. For example ym.master=10.0.2.2 will make the minion to search the master at the address 10.0.2.2.

An internal systemd service in the image will detect this address and configure the salt-minion accordingly.

Under the current Yomi states, this address will be copied under the new installed system, together with the key delivered by the salt-master service. This means that once the system is fully installed with the new operating system, the new salt-minion will find the master directly after the first boot.

In a similar way, during the boot process we can set the minion ID that will be assigned to the salt-minion. Using the parameter ym.minion_id. For example, ym.minion_id=worker01 will set the minion ID for this system as worker01.

The rules for the minion ID are a bit more complicated. Salt, by default, set the minion ID equal to the FQDN or the IP of the node if no ID is specified. This cannot be a good idea if the IP changes, so the current rules are:

• The value from ym.minion_id boot parameter.
• The FQDN hostname of the system, if is different from localhost.
• The MAC address of the first interface of the system.

Sometimes we need to inject some extra configuration into salt-minion before the service runs. For example, we might need to add some grains, or enable some feature in the salt-minion service running inside the image.

To do that we have two options: we can pass an URL with the content, or we can add the full content as a parameter during the boot process.

To pass an URL we should use ym.config_url parameter. For example, ym.config_url=http://server.com/pub/myconfig.cfg will download the configuration file, and will store it under the default name config_url.cfg in /etc/salt/minion.d. We can set a different name from the default via the parameter ym.config_url_name.

In a similar way we can use the parameter ym.config to declare the full content of the user provided configuration file. You need to use quotes to mark the string and escaped control codes to indicate new lines or tabs, like ym.config="grains:\n my_grain: my_value". This will create a file named config.cfg, and the name can be overwritten with the parameter ym.config_name.

Because the versatility of Salt, it's possible to execute the modules that belong to the salt-minion service Yomi without the requirement of any salt-master nor salt-minion service running. We could launch the installation via only the salt-call command in local mode.

Because of that, it is possible to deliver Yomi as a single container, composed of the different Salt and Yomi modules and states.

We can boot a machine using any mechanism, like a recovery image, and use podman to register the Yomi container. This container will be executed as a privileged one, mapping the external devices inside the container space.

To register the container we can do:

podman pull registry.opensuse.org/systemsmanagement/yomi/images/opensuse/yomi:latest

Is recommended to create a local pillar directory;

mkdir pillar

Once we have the pillar data, we can launch the installer:

podman run --privileged --rm \
  -v /dev:/dev \
  -v /run/udev:/run/udev \
  -v ./pillar:/srv/pillar \
  <CONTAINER_ID> \
  salt-call --local state.highstate

Once salt-master is configured and running, the yomi-formula states are available and a new system is booted with a up-to-date salt-minion, we can start to operate with Yomi.

The usual process is simple: describe the pillar information and apply the yomi state to the node or nodes. Is not relevant how the pillar was designed (maybe using a smart template that cover all the cases or writing a raw YAML that only covers one single installation). In this section we will provide some hints about how get information and can help in this process.

The provided pillar are only an example of what we can do with Yomi. Eventually we need to adapt them based on the hardware that we have.

We can discover the hardware configuration with different mechanism. One is get the grains information directly from the minion:

salt node grains.items

We can get more detailed information using other Salt modules, like partition.list, network.interfaces or udev.info.

With Yomi we provided a simple interface to hwinfo that provides some of the information that is required to make decisions about the pillar in a single report:

# Synchronize all the modules to the minion
salt node saltutil.sync_all

# Get a short report about some devices
salt node devices.hwinfo

# Get a detailled report about some devices
salt node devices.hwinfo short=no

The package yomi-formula provides some pillar examples that can be used as a reference when you are creating your own profiles.

Salt search the pillar information in the directories listed in the pillar_roots configuration entry, and using the snippet from the section Pillar in Yomi, we can make those examples available in our system.

In the case that we want to edit those files, we can copy them in a different directory and add it to the pillar_roots entry.

mkdir -p /srv/pillar-yomi
cp -a /usr/share/yomi/pillar/* /srv/pillar-yomi

cat <<EOF > /etc/salt/master.d/pillar.conf
pillar_roots:
  base:
    - /srv/pillar-yomi
    - /srv/pillar
EOF
systemctl restart salt-master.service

The pillar tree start with the top.sls file (there is another top.sls file for the states, do not confuse them).

base:
  '*':
    - installer

This file is used to map the node with the data that the states will use later. For this example the file that contain the data is installer.sls, but feel free to choose a different name when you are creating your own pillar.

This installer.sls is used as an entry point for the rest of the data. Inside the file there is some Jinja templates that can be edited to define different kinds of installations. This feature is leveraged by the openQA tests, to easily make multiple deployments.

You can edit the {% set VAR=VAL %} section to adjust it to your current profile, or create one from scratch. The files _storage.sls.* are included for different scenarios, and this is the place where the disk layout is described. Feel free to include it directly on your pillar, or use a different mechanism to decide the layout.

Yomi tries to be careful with the current data stored on the disks. By default, it will not remove any partition nor will make an implicit decision about the device where the installation will run.

If we want to remove the data from the device, we can use the provided devices.wipe execution module.

# List the partitions
salt node partition.list /dev/sda

# Make sure that the new modules are in the minion
salt node saltutil.sync_all

# Remove all the partitions and the filesystem information
salt node devices.wipe /dev/sda

To wipe all the devices defined in the pillar at once, we can apply the yomi.storage.wipe state.

# Make sure that the new modules are in the minion
salt node saltutil.sync_all

# Remove all the partitions and the filesystem information
salt node state.apply yomi.storage.wipe

Finally, to install the operating system defined by the pillar into the new node, we need to apply the high-state:

salt node state.apply yomi

If we have a top.sls file similar to this example, living in /srv/salt or in any other place where file_roots option is configured:

base:
  '*':
    - yomi

We can apply directly the high state:

salt node state.highstate

To install a new node, we need to provide some data to describe the installation requirements, like the layout of the partitions, file systems used, or what software to install inside the new deployment. This data is collected in what is Salt is known as a pillar.

To configure the salt-master service to find the pillar, check the section Looking for the pillar.

Pillar can be associated with certain nodes in our network, making of this technique a basic one to map a description of how and what to install into a node. This mapping is done via the top.sls file:

base:
  'C7:7E:55:62:83:17':
    - installer

In installer.sls we will describe in detail the installation parameters that will be applied to the node which minion-id match with C7:7E:55:62:83:17. Note that in this example we are using the MAC address of the first interface as a minion-id (check the section Enabling Autosign for an example).

The installer.sls pillar consist on several sections, that we can describe here.

The config section contains global configuration options that will affect the installer.

• events: Boolean. Optional. Default: yes

  Yomi can fire Salt events before and after the execution of the internal states that Yomi use to drive the installation. Using the Salt API, WebSockets, or any other mechanism provided by Salt, we can listen the event bus and use this information to monitor the installer. Yomi provides a basic tool, yomi-monitor, that shows real time information about the installation process.

  To disable the events, set this parameter to no.

  Note that this option will add three new states for each single Yomi state. One extra state is executed always before the normal state, and is used to signalize that a new state will be executed. If the state is successfully terminated, a second extra state will send an event to signalize that the status of the state is positive. But if the state fails, a third state will send the fail signal. All those extra states will be showed in the final report of Salt.

• reboot: String. Optional. Default: yes

  Control the way that the node will reboot. There are three possible values:

  • yes: Will produce a full reboot cycle. This value can be specified as the "yes" string, or the True boolean value.

  • no: Will no reboot after the installation.

  • kexec: Instead of rebooting, reload the new kernel installed in the node.

  • halt: The machine will halt at the end of the installation.

  • shutdown: The machine will shut down at the end of the installation.

• snapper: Boolean. Optional. Default: no

  In Btrfs configurations (and in LVM, but still not implemented) we can install the snapper tool, to do automatic snapshots before and after updates in the system. One installed, a first snapshot will be done and the GRUB entry to boot from snapshots will be added.

• locale: String. Optional. Default: en_US.utf8

  Sets the system locale, more specifically the LANG= and LC_MESSAGES settings. The argument should be a valid locale identifier, such as de_DE.UTF-8. This controls the locale.conf configuration file.

• locale_message: String. Optional.

  Sets the system locale, more specifically the LANG= and LC_MESSAGES settings. The argument should be a valid locale identifier, such as de_DE.UTF-8. This controls the locale.conf configuration file.

• keymap: String. Optional. Default: us

  Sets the system keyboard layout. The argument should be a valid keyboard map, such as de-latin1. This controls the "KEYMAP" entry in the vconsole.conf configuration file.

• timezone: String. Optional. Default: UTC

  Sets the system time zone. The argument should be a valid time zone identifier, such as "Europe/Berlin". This controls the localtime symlink.

• hostname: String. Optional.

  Sets the system hostname. The argument should be a host name, compatible with DNS. This controls the hostname configuration file.

• machine_id: String. Optional.

  Sets the system's machine ID. This controls the machine-id file. If no one is provided, the one from the current system will be re-used.

• target: String. Optional. Default: multi-user.target

  Set the default target used for the boot process.

Example:

config:
  # Do not send events, useful for debugging
  events: no
  # Do not reboot after installation
  reboot: no
  # Always install snapper if possible
  snapper: yes
  # Set language to English / US
  locale: en_US.UTF-8
  # Japanese keyboard
  keymap: jp
  # Universal Timezone
  timezone: UTC
  # Boot in graphical mode
  target: graphical.target

Yomi separate partitioning the devices from providing a file system, creating volumes or building arrays of disks. The advantage of this is that this, usually, compose better that other approaches, and makes more easy adding more options that needs to work correctly with the rest of the system.

• config: Dictionary. Optional.

  Subsection that store some configuration options related with the partitioner.

  • label: String. Optional. Default: msdos

    Default label for the partitions of the devices. We use any parted partition recognized by mklabel, like gpt, msdos or bsd. For UEFI systems, we need to set it to gpt. This value will be used for all the devices if is not overwritten.

  • initial_gap: Integer. Optional. Default: 0

    Initial gap (empty space) leaved before the first partition. Usually is recommended to be 1MB, so GRUB have room to write the code needed after the MBR, and the sectors are aligned for multiple SSD and hard disk devices. Also is relevant for the sector alignment in devices. The valid units are the same for parted. This value will be used for all the devices if is not overwritten.

• devices: Dictionary.

  List of devices that will be partitioned. We can indicate already present devices, like /dev/sda or /dev/hda, but we can also indicate devices that will be present after the RAID configuration, like /dev/md0 or /dev/md/myraid. We can use any valid device name in Linux such as all the /dev/disk/by-id/..., /dev/disk/by-label/..., /dev/disk/by-uuid/... and others.

  For each device we have:

  • label: String. Optional. Default: msdos

    Partition label for the device. The meaning and the possible values are identical for label in the config section.

  • initial_gap: Integer. Optional. Default: 0

    Initial gap (empty space) leave before the first partition for this device.

  • partitions: Array. Optional.

    Partitions inside a device are described with an array. Each element of the array is a dictionary that describe a single partition.

    • number: Integer. Optional. Default: loop.index

      Expected partition number. Eventually this parameter will be really optional, when the partitioner can deduce it from other parameters. Today is better to be explicit in the partition number, as this will guarantee that the partition is found in the hard disk if present. If is not set, number will be the current index position in the array.

    • id: String. Optional.

      Full name of the partition. For example, valid ids can be /dev/sda1, /dev/md0p1, etc. Is optional, as the name can be deduced from number.

    • size: Float or String.

      Size of the partition expressed in parted units. All the units needs to match for partitions on the same device. For example, if initial_gap or the first partition is expressed in MB, all the sized needs to be expressed in MB too.

      The last partition can use the string rest to indicate that this partition will use all the free space available. If after this another partition is defined, Yomi will show a validation error.

    • type: String.

      A string that indicate for what this partition will be used. Yomi recognize several types:

      • swap: This partition will be used for SWAP.
      • linux: Partition used to root, home or any data.
      • boot: Small partition used for GRUB when in BIOS and gpt.
      • efi: EFI partition used by GRUB when UEFI.
      • lvm: Partition used to build an LVM physical volume.
      • raid: Partition that will be a component of an array.

Example:

partitions:
  config:
    label: gpt
    initial_gap: 1MB
  devices:
    /dev/sda:
      partitions:
        - number: 1
          size: 256MB
          type: efi
        - number: 2
          size: 1024MB
          type: swap
        - number: 3
          size: rest
          type: linux

To build an LVM we usually create some partitions (in the partitions section) with the lvm type set, and in the lvm section we describe the details. This section is a dictionary, were each key is the name of the LVM volume, and inside it we can find:

• devices: Array.

  List of components (partitions or full devices) that will constitute the physical volumes and the virtual group of the LVM. If the element of the array is a string, this will be the name of a device (or partition) that belongs to the physical group. If the element is a dictionary it will contains:

  • name: String.

    Name of the device or partition.

  The rest of the elements of the dictionary will be passed to the pvcreate command.

  Note that the name of the virtual group will be the key where this definition is under.

• volumes: Array.

  Each element of the array will define:

  • name: String.

    Name of the logical volume under the volume group.

  The rest of the elements of the dictionary will be passed to the lvcreate command. For example, size and extents are used to indicate the size of the volume, and they can include a suffix to indicate the units. Those units will be the same used for lvcreate.

The rest of the elements of this section will be passed to the vgcreate command.

Example:

lvm:
  system:
    devices:
      - /dev/sda1
      - /dev/sdb1
      - name: /dev/sdc1
        dataalignmentoffset: 7s
    clustered: 'n'
    volumes:
      - name: swap
        size: 1024M
      - name: root
        size: 16384M
      - name: home
        extents: 100%FREE

In the same way that LVM, to create RAID arrays we can setup first partitions (with the type raid) and configure the details in this section. Also, similar to the LVM section, the keys a correspond to the name of the device where the RAID will be created. Valid values are like /dev/md0 or /dev/md/system.

• level: String.

  RAID level. Valid values can be linear, raid0, 0, stripe, raid1, 1, mirror, raid4, 4, raid5, 5, raid6, 6, raid10, 10, multipath, mp, faulty, container.

• devices: Array.

  List of devices or partitions that build the array.

• metadata: String. Optional. Default: default

  Metadata version for the superblock. Valid values are 0, 0.9, 1, 1.0, 1.1, 1.2, default, ddm, imsm.

The user can specify more parameters that will be passed directly to mdadm, like spare-devices to indicate the number of extra devices in the initial array, or chunk to speficy the chunk size.

Example:

raid:
  /dev/md0:
    level: 1
    devices:
      - /dev/sda1
      - /dev/sdb1
      - /dev/sdc1
    spare-devices: 1
    metadata: 1.0

The partitions, devices or arrays created in previous sections usually requires a file system. This section will simply list the device name and the file system (and properties) that will be applied to it.

• filesystem. String.

  File system to apply in the device. Valid values are swap, linux-swap, bfs, btrfs, xfs, cramfs, ext2, ext3, ext4, minix, msdos, vfat. Technically Salt will search for a command that match mkfs.<filesystem>, so the valid options can be more extensive that the one listed here.

• mountpoint. String.

  Mount point where the device will be registered in fstab.

• fat. Integer. Optional.

  If the file system is vfat we can force the FAT size, like 12, 16 or 32.

• subvolumes. Dictionary.

  For btrfs file systems we can specify more details.

  • prefix. String. Optional.

    btrfs sub-volume name where the rest of the sub-volumes will be under. For example, if we set prefix as @ and we create a sub-volume named var, Yomi will create it as @/var.

  • subvolume. Dictionary.

    • path. String.

      Path name for the sub-volume.

    • copy_on_write. Boolean. Optional. Default: yes

      Value for the copy-on-write option in btrfs.

Example:

filesystems:
  /dev/sda1:
    filesystem: vfat
    mountpoint: /boot/efi
    fat: 32
  /dev/sda2:
    filesystem: swap
  /dev/sda3:
    filesystem: btrfs
    mountpoint: /
    subvolumes:
      prefix: '@'
      subvolume:
        - path: home
        - path: opt
        - path: root
        - path: srv
        - path: tmp
        - path: usr/local
        - path: var
          copy_on_write: no
        - path: boot/grub2/i386-pc
        - path: boot/grub2/x86_64-efi

• device: String.

  Device name where GRUB2 will be installed. Yomi will take care of detecting if is a BIOS or an UEFI setup, and also if Secure-Boot in activated, to install and configure the bootloader (or the shim loader)

• timeout: Integer. Optional. Default: 8

  Value for the GRUB_TIMEOUT parameter.

• kernel: String. Optional. Default: splash=silent quiet

  Line assigned to the GRUB_CMDLINE_LINUX_DEFAULT parameter.

• terminal: String. Optional. Default: gfxterm

  Value for the GRUB_TERMINAL parameter.

  If the value is set to serial, we need to add content to the serial_command parameter.

  If the value is set to console, we can pass the console parameters to the kernel parameter. For example, kernel: splash=silent quiet console=tty0 console=ttyS0,115200

• serial_command: String. Optional

  Value for the GRUB_SERIAL_COMMAND parameter. If there is a value, GRUB_TERMINAL is expected to be serial.

• gfxmode: String. Optional. Default: auto

  Value for the GRUB_GFXMODE parameter.

• theme: Boolean. Optional. Default: no

  If yes the grub2-branding package will be installed and configured.

• disable_os_prober: Boolean. Optional. Default: False

  Value for the GRUB_DISABLE_OS_PROBER parameter.

Example:

bootloader:
  device: /dev/sda

We can indicate the repositories that will be registered in the new installation, and the packages and patterns that will be installed.

• config. Dictionary. Optional

  Local configuration for the software section. Except minimal, transfer, and verify all the options can be overwritten in each repository definition.

  • minimal: Boolean. Optional. Default: no

    Configure zypper to make a minimal installation, excluding recommended, documentation and multi-version packages.

  • transfer: Boolean. Optional. Default: no

    Transfer the current repositories (maybe defined in the media installation) into the installed system. If marked, this step will be done early, so any future action could update or replace one of the repositories.

  • verify: Boolean. Optional. Default: yes

    Verify the package key when installing.

  • enabled: Boolean. Optional. Default: yes

    If the repository is enabled, packages can be installed from there. A disabled repository will not be removed.

  • refresh: Boolean. Optional. Default: yes

    Enable auto-refresh of the repository.

  • gpgcheck: Boolean. Optional. Default: yes

    Enable or disable the GPG check for the repositories.

  • gpgautoimport: Boolean. Optional. Default: yes

    If enabled, automatically trust and import public GPG key for the repository.

  • cache: Boolean. Optional. Default: no

    If the cache is enabled, will keep the RPM packages.

• repositories. Dictionary. Optional

  Each key of the dictionary will be the alias under where this repository is registered, and the key, if is a string, the URL associated with it.

  If the key is an dictionary, we can overwrite some of the default configuration options set in the config section, with the exception of minimal. There are some more elements that we can set for the repository:

  • url: String.

    URL of the repository.

  • name: String. Optional

    Descriptive name for the repository.

  • priority: Integer. Optional. Default: 0

    Set priority of the repository.

• packages. Array. Optional

  List of packages or patters to be installed.

• image. Dictionary. Optional

  We can bootstrap the root file system based on a partition image generate by KIWI (or any other mechanism), that will be copied into the partition that have the root mount point assigned. This can be used to speed the installation process.

  Those images needs to contain only the file system and the data. If the image contains a boot loader or partition information, the image will fail during the resize operation. To validate if the image is suitable, a simple file image.raw will do.

  • url: String.

    URL of the image. As internally we are using curl to fetch the image, we can support multiple protocols like http://, https:// or tftp:// among others. The image can be compressed, and in that case one of those extensions must to be used to indicate the format: [gz, bz2, xz]

  • md5|sha1|sha224|sha256|sha384|sha512: String. Optional

    Checksum type and value used to validate the image. If this field is present but empty (only the checksum type, but with no value attached), the state will try to fetch the checksum fail from the same URL given in the previous field. If the path contains an extension for a compression format, this will be replaced with the checksum type as a new extension.

    For example, if the URL is http://example.com/image.xz, the checksum type is md5, and no value is provided, the checksum will be expected at http://example.com/image.md5.

    But if the URL is something like http://example.com/image.ext4, the checksum will be expected in the URL http://example.com/image.ext4.md5.

  If the checksum type is provided, the value for the last image will be stored in the Salt cache, and will be used to decide if the image in the URL is different from the one already copied in the partition. If this is the case, no image will be downloaded. Otherwise a new image will be copied, and the old one will be overwritten in the same partition.

Example:

software:
  repositories:
    repo-oss: "http://download.opensuse.org/tumbleweed/repo/oss"
    update:
	  url: http://download.opensuse.org/update/tumbleweed/
	  name: openSUSE Update
  packages:
    - patterns-base-base
    - kernel-default

Very related with the previous section (software), we can register an SLE product and modules using the SUSEConnect command.

In order to SUSEConnect to succeed, a product needs to be present already in the system. This imply that the register must happen after (at least a partial) installation has been done.

As SUSEConnect will register new repositories, this also imply that not all the packages that can be enumerated in the software section can be installed.

To resolve both conflicts, Yomi will first install the packages listed in the sofwtare section, and after the registration, the packages listed in this suseconnect section.

• config. Dictionary.

  Local configuration for the section. It is not optional as there is at least one parameter that is required for any registration.

  • regcode. String.

  Subscription registration code for the product to be registered.

  • email. String. Optional.

  Email address for product registration.

  • url. String. Optional.

  URL of registration server (e.g. https://scc.suse.com)

  • version. String. Optional.

  Version part of the product name. If the product name do not have a version, this default value will be used.

  • arch. String. Optional.

  Architecture part of the product name. If the product name do not have an architecture, this default value will be used.

• products. Array. Optional.

  Product names to register. The expected format is //. If only is used, the values for and will be taken from the config section.

  If the product / module have a different registration code than the one declared in the config sub-section, we can declare a new one via a dictionary.

  • name. String. Optional.

    Product names to register. The expected format is //. If only is used, the values for and will be taken from the config section.

  • regcode. String. Optional.

    Subscription registration code for the product to be registered.

• packages. Array. Optional

  List of packages or patters to be installed from the different modules.

Example:

suseconnect:
  config:
    regcode: SECRET-CODE
  products:
    - sle-module-basesystem/15.2/x86_64
    - sle-module-server-applications/15.2/x86_64
    - name: sle-module-live-patching/15.2/x86_64
      regcode: SECRET-CODE

Install and configure the salt-minion service.

• config. Boolean. Optional. Default: no

  If yes, the configuration and cetificates of the new minion will be the same that the current minion that is activated. This will copy the minion configuration, certificates and grains, together with the cached modules and states that are usually synchronized before a highstate.

  This option will be replaced in the future with more detailed ones.

Example:

salt-minion:
  config: yes

We can list the services that will be enabled or disabled during boot time.

• enabled. Array. Optional

  List of services that will be enabled and started during the boot.

• disabled. Array. Optional

  List of services that will be exclicitly disabled during the boot.

Example:

services:
  enabled:
    - salt-minion

We can list the networks available in the target system. If the list is not provided, Yomi will try to deduce the network configuration based on the current setup.

• interface. String.

  Name of the interface.

Example:

networks:
  - interface: ens3

In this section we can list a simple list of users and passwords that we expect to find once the system is booted.

• username. String.

  Login or username for the user.

• password. String. Optional.

  Shadow password hash for the user.

• certificates. Array. Optional.

  Certificates that will be added to .ssh/authorized_keys. Use only the encoded key (remove the "ssh-rsa" prefix and the "user@host" suffix).

Example:

users:
  - username: root
    password: "$1$wYJUgpM5$RXMMeASDc035eX.NbYWFl0"
  - username: aplanas
    certificates:
      - "AAAAB3NzaC1yc2EAAAADAQABAAABAQDdP6oez825gnOLVZu70KqJXpqL4fGf\
        aFNk87GSk3xLRjixGtr013+hcN03ZRKU0/2S7J0T/dICc2dhG9xAqa/A31Qac\
        hQeg2RhPxM2SL+wgzx0geDmf6XDhhe8reos5jgzw6Pq59gyWfurlZaMEZAoOY\
        kfNb5OG4vQQN8Z7hldx+DBANPbylApurVz6h5vvRrkPfuRVN5ZxOkI+LeWhpo\
        vX5XK3eTjetAwWEro6AAXpGoQQQDjSOoYHCUmXzcZkmIWEubCZvAI4RZ+XCZs\
        +wTeO2RIRsunqP8J+XW4cZ28RZBc9K4I1BV8C6wBxN328LRQcilzw+Me+Lfre\
        eDPglqx"