1 - Configuration Patches

In this guide, we’ll patch the generated machine configuration.

Talos generates machine configuration for two types of machines: controlplane and worker machines. Many configuration options can be adjusted using talosctl gen config but not all of them. Configuration patching allows modifying machine configuration to fit it for the cluster or a specific machine.

Configuration Patch Formats

Talos supports two configuration patch formats:

  • strategic merge patches
  • RFC6902 (JSON patches)

Strategic merge patches are the easiest to use, but JSON patches allow more precise configuration adjustments.

Note: Talos 1.5+ supports multi-document machine configuration. JSON patches don’t support multi-document machine configuration, while strategic merge patches do.

Strategic Merge patches

Strategic merge patches look like incomplete machine configuration files:

machine:
  network:
    hostname: worker1

When applied to the machine configuration, the patch gets merged with the respective section of the machine configuration:

machine:
  network:
    interfaces:
      - interface: eth0
        addresses:
          - 10.0.0.2/24
    hostname: worker1

In general, machine configuration contents are merged with the contents of the strategic merge patch, with strategic merge patch values overriding machine configuration values. There are some special rules:

  • If the field value is a list, the patch value is appended to the list, with the following exceptions:
    • values of the fields cluster.network.podSubnets and cluster.network.serviceSubnets are overwritten on merge
    • network.interfaces section is merged with the value in the machine config if there is a match on interface: or deviceSelector: keys
    • network.interfaces.vlans section is merged with the value in the machine config if there is a match on the vlanId: key
    • cluster.apiServer.auditPolicy value is replaced on merge
    • ExtensionServiceConfig.configFiles section is merged matching on mountPath (replacing content if matches)

When patching a multi-document machine configuration, following rules apply:

  • for each document in the patch, the document is merged with the respective document in the machine configuration (matching by kind, apiVersion and name for named documents)
  • if the patch document doesn’t exist in the machine configuration, it is appended to the machine configuration

The strategic merge patch itself might be a multi-document YAML, and each document will be applied as a patch to the base machine configuration. Keep in mind that you can’t patch the same document multiple times with the same patch.

You can also delete parts from the configuration using $patch: delete syntax similar to the Kubernetes strategic merge patch.

For example, with configuration:

machine:
  network:
    interfaces:
      - interface: eth0
        addresses:
          - 10.0.0.2/24
    hostname: worker1

and patch document:

machine:
  network:
    interfaces:
    - interface: eth0
      $patch: delete
    hostname: worker1

The resulting configuration will be:

machine:
  network:
    hostname: worker1

You can also delete entire docs (but not the main v1alpha1 configuration!) using this syntax:

apiVersion: v1alpha1
kind: SideroLinkConfig
$patch: delete
---
apiVersion: v1alpha1
kind: ExtensionServiceConfig
name: foo
$patch: delete

This will remove the documents SideroLinkConfig and ExtensionServiceConfig with name foo from the configuration.

RFC6902 (JSON Patches)

JSON patches can be written either in JSON or YAML format. A proper JSON patch requires an op field that depends on the machine configuration contents: whether the path already exists or not.

For example, the strategic merge patch from the previous section can be written either as:

- op: replace
  path: /machine/network/hostname
  value: worker1

or:

- op: add
  path: /machine/network/hostname
  value: worker1

The correct op depends on whether the /machine/network/hostname section exists already in the machine config or not.

Examples

Machine Network

Base machine configuration:

# ...
machine:
  network:
    interfaces:
      - interface: eth0
        dhcp: false
        addresses:
          - 192.168.10.3/24

The goal is to add a virtual IP 192.168.10.50 to the eth0 interface and add another interface eth1 with DHCP enabled.

machine:
  network:
    interfaces:
      - interface: eth0
        vip:
          ip: 192.168.10.50
      - interface: eth1
        dhcp: true
- op: add
  path: /machine/network/interfaces/0/vip
  value:
    ip: 192.168.10.50
- op: add
  path: /machine/network/interfaces/-
  value:
    interface: eth1
    dhcp: true

Patched machine configuration:

machine:
  network:
    interfaces:
      - interface: eth0
        dhcp: false
        addresses:
          - 192.168.10.3/24
        vip:
          ip: 192.168.10.50
      - interface: eth1
        dhcp: true

Cluster Network

Base machine configuration:

cluster:
  network:
    dnsDomain: cluster.local
    podSubnets:
      - 10.244.0.0/16
    serviceSubnets:
      - 10.96.0.0/12

The goal is to update pod and service subnets and disable default CNI (Flannel).

cluster:
  network:
    podSubnets:
      - 192.168.0.0/16
    serviceSubnets:
      - 192.0.0.0/12
    cni:
      name: none
- op: replace
  path: /cluster/network/podSubnets
  value:
    - 192.168.0.0/16
- op: replace
  path: /cluster/network/serviceSubnets
  value:
    - 192.0.0.0/12
- op: add
  path: /cluster/network/cni
  value:
    name: none

Patched machine configuration:

cluster:
  network:
    dnsDomain: cluster.local
    podSubnets:
      - 192.168.0.0/16
    serviceSubnets:
      - 192.0.0.0/12
    cni:
      name: none

Kubelet

Base machine configuration:

# ...
machine:
  kubelet: {}

The goal is to set the kubelet node IP to come from the subnet 192.168.10.0/24.

machine:
  kubelet:
    nodeIP:
      validSubnets:
        - 192.168.10.0/24
- op: add
  path: /machine/kubelet/nodeIP
  value:
    validSubnets:
      - 192.168.10.0/24

Patched machine configuration:

machine:
  kubelet:
    nodeIP:
      validSubnets:
        - 192.168.10.0/24

Admission Control: Pod Security Policy

Base machine configuration:

cluster:
  apiServer:
    admissionControl:
      - name: PodSecurity
        configuration:
          apiVersion: pod-security.admission.config.k8s.io/v1alpha1
          defaults:
            audit: restricted
            audit-version: latest
            enforce: baseline
            enforce-version: latest
            warn: restricted
            warn-version: latest
          exemptions:
            namespaces:
              - kube-system
            runtimeClasses: []
            usernames: []
          kind: PodSecurityConfiguration

The goal is to add an exemption for the namespace rook-ceph.

cluster:
  apiServer:
    admissionControl:
      - name: PodSecurity
        configuration:
          exemptions:
            namespaces:
              - rook-ceph
- op: add
  path: /cluster/apiServer/admissionControl/0/configuration/exemptions/namespaces/-
  value: rook-ceph

Patched machine configuration:

cluster:
  apiServer:
    admissionControl:
      - name: PodSecurity
        configuration:
          apiVersion: pod-security.admission.config.k8s.io/v1alpha1
          defaults:
            audit: restricted
            audit-version: latest
            enforce: baseline
            enforce-version: latest
            warn: restricted
            warn-version: latest
          exemptions:
            namespaces:
              - kube-system
              - rook-ceph
            runtimeClasses: []
            usernames: []
          kind: PodSecurityConfiguration

Configuration Patching with talosctl CLI

Several talosctl commands accept config patches as command-line flags. Config patches might be passed either as an inline value or as a reference to a file with @file.patch syntax:

talosctl ... --patch '[{"op": "add", "path": "/machine/network/hostname", "value": "worker1"}]' --patch @file.patch

If multiple config patches are specified, they are applied in the order of appearance. The format of the patch (JSON patch or strategic merge patch) is detected automatically.

Talos machine configuration can be patched at the moment of generation with talosctl gen config:

talosctl gen config test-cluster https://172.20.0.1:6443 --config-patch @all.yaml --config-patch-control-plane @cp.yaml --config-patch-worker @worker.yaml

Generated machine configuration can also be patched after the fact with talosctl machineconfig patch

talosctl machineconfig patch worker.yaml --patch @patch.yaml -o worker1.yaml

Machine configuration on the running Talos node can be patched with talosctl patch:

talosctl patch mc --nodes 172.20.0.2 --patch @patch.yaml

2 - Containerd

Customize Containerd Settings

The base containerd configuration expects to merge in any additional configs present in /etc/cri/conf.d/20-customization.part.

Examples

Exposing Metrics

Patch the machine config by adding the following:

machine:
  files:
    - content: |
        [metrics]
          address = "0.0.0.0:11234"        
      path: /etc/cri/conf.d/20-customization.part
      op: create

Once the server reboots, metrics are now available:

$ curl ${IP}:11234/v1/metrics
# HELP container_blkio_io_service_bytes_recursive_bytes The blkio io service bytes recursive
# TYPE container_blkio_io_service_bytes_recursive_bytes gauge
container_blkio_io_service_bytes_recursive_bytes{container_id="0677d73196f5f4be1d408aab1c4125cf9e6c458a4bea39e590ac779709ffbe14",device="/dev/dm-0",major="253",minor="0",namespace="k8s.io",op="Async"} 0
container_blkio_io_service_bytes_recursive_bytes{container_id="0677d73196f5f4be1d408aab1c4125cf9e6c458a4bea39e590ac779709ffbe14",device="/dev/dm-0",major="253",minor="0",namespace="k8s.io",op="Discard"} 0
...
...

Pause Image

This change is often required for air-gapped environments, as containerd CRI plugin has a reference to the pause image which is used to create pods, and it can’t be controlled with Kubernetes pod definitions.

machine:
  files:
    - content: |
        [plugins]
          [plugins."io.containerd.cri.v1.images".pinned_images]
            sandbox = "registry.k8s.io/pause:3.8"        
      path: /etc/cri/conf.d/20-customization.part
      op: create

Now the pause image is set to registry.k8s.io/pause:3.8:

$ talosctl containers --kubernetes
NODE         NAMESPACE   ID                                                              IMAGE                                                      PID    STATUS
172.20.0.5   k8s.io      kube-system/kube-flannel-6hfck                                  registry.k8s.io/pause:3.8                                  1773   SANDBOX_READY
172.20.0.5   k8s.io      └─ kube-system/kube-flannel-6hfck:install-cni:bc39fec3cbac      ghcr.io/siderolabs/install-cni:v1.3.0-alpha.0-2-gb155fa0   0      CONTAINER_EXITED
172.20.0.5   k8s.io      └─ kube-system/kube-flannel-6hfck:install-config:5c3989353b98   ghcr.io/siderolabs/flannel:v0.20.1                         0      CONTAINER_EXITED
172.20.0.5   k8s.io      └─ kube-system/kube-flannel-6hfck:kube-flannel:116c67b50da8     ghcr.io/siderolabs/flannel:v0.20.1                         2092   CONTAINER_RUNNING
172.20.0.5   k8s.io      kube-system/kube-proxy-xp7jq                                    registry.k8s.io/pause:3.8                                  1780   SANDBOX_READY
172.20.0.5   k8s.io      └─ kube-system/kube-proxy-xp7jq:kube-proxy:84fc77c59e17         registry.k8s.io/kube-proxy:v1.26.0-alpha.3                 1843   CONTAINER_RUNNING

3 - Custom Certificate Authorities

How to supply custom certificate authorities

Appending the Certificate Authority

Append additional certificate authorities to the system’s trusted certificate store by patching the machine configuration with the following document:

apiVersion: v1alpha1
kind: TrustedRootsConfig
name: custom-ca
certificates: |-
    -----BEGIN CERTIFICATE-----
    ...
    -----END CERTIFICATE-----    

Multiple documents can be appended, and multiple CA certificates might be present in each configuration document.

This configuration can be also applied in maintenance mode.

4 - Disk Encryption

Guide on using system disk encryption

It is possible to enable encryption for system disks at the OS level. Currently, only STATE and EPHEMERAL partitions can be encrypted. STATE contains the most sensitive node data: secrets and certs. The EPHEMERAL partition may contain sensitive workload data. Data is encrypted using LUKS2, which is provided by the Linux kernel modules and cryptsetup utility. The operating system will run additional setup steps when encryption is enabled.

If the disk encryption is enabled for the STATE partition, the system will:

  • Save STATE encryption config as JSON in the META partition.
  • Before mounting the STATE partition, load encryption configs either from the machine config or from the META partition. Note that the machine config is always preferred over the META one.
  • Before mounting the STATE partition, format and encrypt it. This occurs only if the STATE partition is empty and has no filesystem.

If the disk encryption is enabled for the EPHEMERAL partition, the system will:

  • Get the encryption config from the machine config.
  • Before mounting the EPHEMERAL partition, encrypt and format it.

This occurs only if the EPHEMERAL partition is empty and has no filesystem.

Talos Linux supports four encryption methods, which can be combined together for a single partition:

  • static - encrypt with the static passphrase (weakest protection, for STATE partition encryption it means that the passphrase will be stored in the META partition).
  • nodeID - encrypt with the key derived from the node UUID (weak, it is designed to protect against data being leaked or recovered from a drive that has been removed from a Talos Linux node).
  • kms - encrypt using key sealed with network KMS (strong, but requires network access to decrypt the data.)
  • tpm - encrypt with the key derived from the TPM (strong, when used with SecureBoot).

Note: nodeID encryption is not designed to protect against attacks where physical access to the machine, including the drive, is available. It uses the hardware characteristics of the machine in order to decrypt the data, so drives that have been removed, or recycled from a cloud environment or attached to a different virtual machine, will maintain their protection and encryption.

Configuration

Disk encryption is disabled by default. To enable disk encryption you should modify the machine configuration with the following options:

machine:
  ...
  systemDiskEncryption:
    ephemeral:
      provider: luks2
      keys:
        - nodeID: {}
          slot: 0
    state:
      provider: luks2
      keys:
        - nodeID: {}
          slot: 0

Encryption Keys

Note: What the LUKS2 docs call “keys” are, in reality, a passphrase. When this passphrase is added, LUKS2 runs argon2 to create an actual key from that passphrase.

LUKS2 supports up to 32 encryption keys and it is possible to specify all of them in the machine configuration. Talos always tries to sync the keys list defined in the machine config with the actual keys defined for the LUKS2 partition. So if you update the keys list, keep at least one key that is not changed to be used for key management.

When you define a key you should specify the key kind and the slot:

machine:
  ...
  state:
    keys:
      - nodeID: {} # key kind
        slot: 1

  ephemeral:
    keys:
      - static:
          passphrase: supersecret
        slot: 0

Take a note that key order does not play any role on which key slot is used. Every key must always have a slot defined.

Encryption Key Kinds

Talos supports two kinds of keys:

  • nodeID which is generated using the node UUID and the partition label (note that if the node UUID is not really random it will fail the entropy check).
  • static which you define right in the configuration.
  • kms which is sealed with the network KMS.
  • tpm which is sealed using the TPM and protected with SecureBoot.

Note: Use static keys only if your STATE partition is encrypted and only for the EPHEMERAL partition. For the STATE partition it will be stored in the META partition, which is not encrypted.

Key Rotation

In order to completely rotate keys, it is necessary to do talosctl apply-config a couple of times, since there is a need to always maintain a single working key while changing the other keys around it.

So, for example, first add a new key:

machine:
  ...
  ephemeral:
    keys:
      - static:
          passphrase: oldkey
        slot: 0
      - static:
          passphrase: newkey
        slot: 1
  ...

Run:

talosctl apply-config -n <node> -f config.yaml

Then remove the old key:

machine:
  ...
  ephemeral:
    keys:
      - static:
          passphrase: newkey
        slot: 1
  ...

Run:

talosctl apply-config -n <node> -f config.yaml

Going from Unencrypted to Encrypted and Vice Versa

Ephemeral Partition

There is no in-place encryption support for the partitions right now, so to avoid losing data only empty partitions can be encrypted.

As such, migration from unencrypted to encrypted needs some additional handling, especially around explicitly wiping partitions.

  • apply-config should be called with --mode=staged.
  • Partition should be wiped after apply-config, but before the reboot.

Edit your machine config and add the encryption configuration:

vim config.yaml

Apply the configuration with --mode=staged:

talosctl apply-config -f config.yaml -n <node ip> --mode=staged

Wipe the partition you’re going to encrypt:

talosctl reset --system-labels-to-wipe EPHEMERAL -n <node ip> --reboot=true

That’s it! After you run the last command, the partition will be wiped and the node will reboot. During the next boot the system will encrypt the partition.

State Partition

Calling wipe against the STATE partition will make the node lose the config, so the previous flow is not going to work.

The flow should be to first wipe the STATE partition:

talosctl reset  --system-labels-to-wipe STATE -n <node ip> --reboot=true

Node will enter into maintenance mode, then run apply-config with --insecure flag:

talosctl apply-config --insecure -n <node ip> -f config.yaml

After installation is complete the node should encrypt the STATE partition.

5 - Disk Management

Guide on managing disks

Talos Linux version 1.8.0 introduces a new backend for managing system and user disks. The machine configuration changes required are minimal, and the new backend is fully compatible with the existing machine configuration.

Listing Disks

To obtain a list of all available block devices (disks) on the machine, you can use the following command:

$ talosctl get disks
NODE         NAMESPACE   TYPE   ID        VERSION   SIZE    READ ONLY   TRANSPORT   ROTATIONAL   WWID                                                               MODEL            SERIAL
172.20.0.5   runtime     Disk   loop0     1         75 MB   true
172.20.0.5   runtime     Disk   nvme0n1   1         10 GB   false       nvme                     nvme.1b36-6465616462656566-51454d55204e564d65204374726c-00000001   QEMU NVMe Ctrl   deadbeef
172.20.0.5   runtime     Disk   sda       1         10 GB   false       virtio      true                                                                            QEMU HARDDISK
172.20.0.5   runtime     Disk   sdb       1         10 GB   false       sata        true         t10.ATA     QEMU HARDDISK                           QM00013        QEMU HARDDISK
172.20.0.5   runtime     Disk   sdc       1         10 GB   false       sata        true         t10.ATA     QEMU HARDDISK                           QM00001        QEMU HARDDISK
172.20.0.5   runtime     Disk   vda       1         13 GB   false       virtio      true

To obtain detailed information about a specific disk, execute the following command:

# talosctl get disk sda -o yaml
node: 172.20.0.5
metadata:
    namespace: runtime
    type: Disks.block.talos.dev
    id: sda
    version: 1
    owner: block.DisksController
    phase: running
    created: 2024-08-29T13:06:43Z
    updated: 2024-08-29T13:06:43Z
spec:
    dev_path: /dev/sda
    size: 10485760000
    human_size: 10 GB
    io_size: 512
    sector_size: 512
    readonly: false
    cdrom: false
    model: QEMU HARDDISK
    modalias: scsi:t-0x00
    bus_path: /pci0000:00/0000:00:07.0/virtio4/host1/target1:0:0/1:0:0:0
    sub_system: /sys/class/block
    transport: virtio
    rotational: true

Discovering Volumes

Talos Linux monitors all block devices and partitions on the machine. Details about these devices, including their type, can be found in the DiscoveredVolume resource.

$ talosctl get discoveredvolumes
NODE         NAMESPACE   TYPE               ID        VERSION   TYPE        SIZE     DISCOVERED   LABEL       PARTITIONLABEL
172.20.0.5   runtime     DiscoveredVolume   dm-0      1         disk        88 MB    xfs          STATE
172.20.0.5   runtime     DiscoveredVolume   loop0     1         disk        75 MB    squashfs
172.20.0.5   runtime     DiscoveredVolume   nvme0n1   1         disk        10 GB
172.20.0.5   runtime     DiscoveredVolume   sda       1         disk        10 GB
172.20.0.5   runtime     DiscoveredVolume   sdb       1         disk        10 GB
172.20.0.5   runtime     DiscoveredVolume   sdc       1         disk        2.1 GB   gpt
172.20.0.5   runtime     DiscoveredVolume   sdc1      1         partition   957 MB   xfs
172.20.0.5   runtime     DiscoveredVolume   sdc2      1         partition   957 MB   xfs
172.20.0.5   runtime     DiscoveredVolume   sdd       1         disk        1.0 GB   gpt
172.20.0.5   runtime     DiscoveredVolume   sdd1      1         partition   957 MB   xfs
172.20.0.5   runtime     DiscoveredVolume   sde       1         disk        10 GB
172.20.0.5   runtime     DiscoveredVolume   vda       1         disk        13 GB    gpt
172.20.0.5   runtime     DiscoveredVolume   vda1      1         partition   105 MB   vfat                     EFI
172.20.0.5   runtime     DiscoveredVolume   vda2      1         partition   1.0 MB                            BIOS
172.20.0.5   runtime     DiscoveredVolume   vda3      1         partition   982 MB   xfs          BOOT        BOOT
172.20.0.5   runtime     DiscoveredVolume   vda4      2         partition   1.0 MB   talosmeta                META
172.20.0.5   runtime     DiscoveredVolume   vda5      1         partition   105 MB   luks                     STATE
172.20.0.5   runtime     DiscoveredVolume   vda6      1         partition   12 GB    xfs          EPHEMERAL   EPHEMERAL

Talos Linux has built-in automatic detection for various filesystem types and GPT partition tables. Currently, the following filesystem types are supported:

  • bluestore (Ceph)
  • ext2, ext3, ext4
  • iso9660
  • luks (LUKS encrypted partition)
  • lvm2
  • squashfs
  • swap
  • talosmeta (Talos Linux META partition)
  • vfat
  • xfs
  • zfs

The discovered volumes can include both Talos-managed volumes and any other volumes present on the machine, such as Ceph volumes.

Volume Management

Talos Linux implements disk management through the concept of volumes. A volume represents a provisioned, located, mounted, or unmounted entity, such as a disk, partition, or tmpfs filesystem.

The configuration of volumes is defined using the VolumeConfig resource, while the current state of volumes is stored in the VolumeStatus resource.

Configuration

The volume configuration is managed by Talos Linux based on machine configuration. To see configured volumes, use the following command:

$ talosctl get volumeconfigs
NODE         NAMESPACE   TYPE           ID                                            VERSION
172.20.0.5   runtime     VolumeConfig   /dev/disk/by-id/ata-QEMU_HARDDISK_QM00001-1   2
172.20.0.5   runtime     VolumeConfig   /dev/disk/by-id/ata-QEMU_HARDDISK_QM00001-2   2
172.20.0.5   runtime     VolumeConfig   /dev/disk/by-id/ata-QEMU_HARDDISK_QM00003-1   2
172.20.0.5   runtime     VolumeConfig   EPHEMERAL                                     2
172.20.0.5   runtime     VolumeConfig   META                                          2
172.20.0.5   runtime     VolumeConfig   STATE                                         4

In the provided output, the volumes EPHEMERAL, META, and STATE are system volumes managed by Talos, while the remaining volumes are based on the machine configuration for machine.disks.

To get details about a specific volume configuration, use the following command:

# talosctl get volumeconfig STATE -o yaml
node: 172.20.0.5
metadata:
    namespace: runtime
    type: VolumeConfigs.block.talos.dev
    id: STATE
    version: 4
    owner: block.VolumeConfigController
    phase: running
    created: 2024-08-29T13:22:04Z
    updated: 2024-08-29T13:22:17Z
    finalizers:
        - block.VolumeManagerController
spec:
    type: partition
    provisioning:
        wave: -1
        diskSelector:
            match: system_disk
        partitionSpec:
            minSize: 104857600
            maxSize: 104857600
            grow: false
            label: STATE
            typeUUID: 0FC63DAF-8483-4772-8E79-3D69D8477DE4
        filesystemSpec:
            type: xfs
            label: STATE
    encryption:
        provider: luks2
        keys:
            - slot: 0
              type: nodeID
    locator:
        match: volume.partition_label == "STATE"
    mount:
        targetPath: /system/state

Status

Current volume status can be obtained using the following command:

$ talosctl get volumestatus
NODE         NAMESPACE   TYPE           ID                                            VERSION   PHASE   LOCATION         SIZE
172.20.0.5   runtime     VolumeStatus   /dev/disk/by-id/ata-QEMU_HARDDISK_QM00001-1   1         ready   /dev/sdc1        957 MB
172.20.0.5   runtime     VolumeStatus   /dev/disk/by-id/ata-QEMU_HARDDISK_QM00001-2   1         ready   /dev/sdc2        957 MB
172.20.0.5   runtime     VolumeStatus   /dev/disk/by-id/ata-QEMU_HARDDISK_QM00003-1   1         ready   /dev/sdd1        957 MB
172.20.0.5   runtime     VolumeStatus   EPHEMERAL                                     1         ready   /dev/nvme0n1p1   10 GB
172.20.0.5   runtime     VolumeStatus   META                                          2         ready   /dev/vda4        524 kB
172.20.0.5   runtime     VolumeStatus   STATE                                         2         ready   /dev/vda5        92 MB

Each volume goes through different phases during its lifecycle:

  • waiting: the volume is waiting to be provisioned
  • missing: all disks have been discovered, but the volume cannot be found
  • located: the volume is found without prior provisioning
  • provisioned: the volume has been provisioned (e.g., partitioned, resized if necessary)
  • prepared: the encrypted volume is open
  • ready: the volume is formatted and ready to be mounted
  • closed: the encrypted volume is closed

Machine Configuration

Note: In Talos Linux 1.8, only EPHEMERAL system volume configuration can be managed through the machine configuration.

Note: The volume configuration in the machine configuration is only applied when the volume has not been provisioned yet. So applying changes after the initial provisioning will not have any effect.

To configure the EPHEMERAL (/var) volume, add the following document to the machine configuration:

apiVersion: v1alpha1
kind: VolumeConfig
name: EPHEMERAL
provisioning:
  diskSelector:
    match: disk.transport == 'nvme'
  minSize: 2GB
  maxSize: 40GB
  grow: false

Every field in the VolumeConfig resource is optional, and if a field is not specified, the default value is used. The default built-in values are:

provisioning:
    diskSelector:
        match: system_disk
    minSize: 2GiB
    grow: true

By default, the EPHEMERAL volume is provisioned on the system disk, which is the disk where Talos Linux is installed. It has a minimum size of 2 GiB and automatically grows to utilize the maximum available space on the disk.

Disk Selector

The diskSelector field is utilized to choose the disk where the volume will be provisioned. It is a Common Expression Language (CEL) expression that evaluates against the available disks. The volume will be provisioned on the first disk that matches the expression and has sufficient free space for the volume.

The expression is evaluated in the following context:

  • system_disk (bool) - indicates if the disk is the system disk
  • disk (Disks.block.talos.dev) - the disk resource being evaluated

For the disk resource, any field available in the resource specification can be used (use talosctl get disks -o yaml to see the output for your machine):

dev_path: /dev/nvme0n1
size: 10485760000
pretty_size: 10 GB
io_size: 512
sector_size: 512
readonly: false
cdrom: false
model: QEMU NVMe Ctrl
serial: deadbeef
wwid: nvme.1b36-6465616462656566-51454d55204e564d65204374726c-00000001
bus_path: /pci0000:00/0000:00:09.0/nvme
sub_system: /sys/class/block
transport: nvme

Additionally, constants for disk size multipliers are available:

  • KiB, MiB, GiB, TiB, PiB, EiB - binary size multipliers (1024)
  • kB, MB, GB, TB, PB, EB - decimal size multipliers (1000)

The disk expression is evaluated against each available disk, and the expression should either return true or false. If the expression returns true, the disk is selected for provisioning.

Note: In CEL, signed and unsigned integers are not interchangeable. Disk sizes are represented as unsigned integers, so suffix u should be used in constants to avoid type mismatch, e.g. disk.size > 10u * GiB.

Examples of disk selector expressions:

  • disk.transport == 'nvme': select the NVMe disks only
  • disk.transport == 'scsi' && disk.size < 2u * TiB: select SCSI disks smaller than 2 TiB
  • disk.serial.startsWith('deadbeef') && !cdrom: select disks with serial number starting with deadbeef and not of CD-ROM type

Minimum and Maximum Size

The minSize and maxSize fields define the minimum and maximum size of the volume, respectively. Talos Linux will always ensure that the volume is at least minSize in size and will not exceed maxSize. If maxSize is not set, the volume will grow to utilize the maximum available space on the disk.

If grow is set to true, the volume will automatically grow to utilize the maximum available space on the disk on each boot.

Setting minSize might influence disk selection - if the disk does not have enough free space to satisfy the minimum size requirement, it will not be selected for provisioning.

6 - Editing Machine Configuration

How to edit and patch Talos machine configuration, with reboot, immediately, or stage update on reboot.

Talos node state is fully defined by machine configuration. Initial configuration is delivered to the node at bootstrap time, but configuration can be updated while the node is running.

There are three talosctl commands which facilitate machine configuration updates:

  • talosctl apply-config to apply configuration from the file
  • talosctl edit machineconfig to launch an editor with existing node configuration, make changes and apply configuration back
  • talosctl patch machineconfig to apply automated machine configuration via JSON patch

Each of these commands can operate in one of four modes:

  • apply change in automatic mode (default): reboot if the change can’t be applied without a reboot, otherwise apply the change immediately
  • apply change with a reboot (--mode=reboot): update configuration, reboot Talos node to apply configuration change
  • apply change immediately (--mode=no-reboot flag): change is applied immediately without a reboot, fails if the change contains any fields that can not be updated without a reboot
  • apply change on next reboot (--mode=staged): change is staged to be applied after a reboot, but node is not rebooted
  • apply change with automatic revert (--mode=try): change is applied immediately (if not possible, returns an error), and reverts it automatically in 1 minute if no configuration update is applied
  • apply change in the interactive mode (--mode=interactive; only for talosctl apply-config): launches TUI based interactive installer

Note: applying change on next reboot (--mode=staged) doesn’t modify current node configuration, so next call to talosctl edit machineconfig --mode=staged will not see changes

Additionally, there is also talosctl get machineconfig -o yaml, which retrieves the current node configuration API resource and contains the machine configuration in the .spec field. It can be used to modify the configuration locally before being applied to the node.

The list of config changes allowed to be applied immediately in Talos v1.8.0:

  • .debug
  • .cluster
  • .machine.time
  • .machine.ca
  • .machine.acceptedCAs
  • .machine.certCANs
  • .machine.install (configuration is only applied during install/upgrade)
  • .machine.network
  • .machine.nodeAnnotations
  • .machine.nodeLabels
  • .machine.nodeTaints
  • .machine.sysfs
  • .machine.sysctls
  • .machine.logging
  • .machine.controlplane
  • .machine.kubelet
  • .machine.pods
  • .machine.kernel
  • .machine.registries (CRI containerd plugin will not pick up the registry authentication settings without a reboot)
  • .machine.features.kubernetesTalosAPIAccess
  • .machine.features.hostDNS
  • .machine.features.kubePrism

talosctl apply-config

This command is traditionally used to submit initial machine configuration generated by talosctl gen config to the node.

It can also be used to apply configuration to running nodes. The initial YAML for this is typically obtained using talosctl get machineconfig -o yaml | yq eval .spec >machs.yaml. (We must use yq because for historical reasons, get returns the configuration as a full resource, while apply-config only accepts the raw machine config directly.)

Example:

talosctl -n <IP> apply-config -f config.yaml

Command apply-config can also be invoked as apply machineconfig:

talosctl -n <IP> apply machineconfig -f config.yaml

Applying machine configuration immediately (without a reboot):

talosctl -n IP apply machineconfig -f config.yaml --mode=no-reboot

Starting the interactive installer:

talosctl -n IP apply machineconfig --mode=interactive

Note: when a Talos node is running in the maintenance mode it’s necessary to provide --insecure (-i) flag to connect to the API and apply the config.

talosctl edit machineconfig

Command talosctl edit loads current machine configuration from the node and launches configured editor to modify the config. If config hasn’t been changed in the editor (or if updated config is empty), update is not applied.

Note: Talos uses environment variables TALOS_EDITOR, EDITOR to pick up the editor preference. If environment variables are missing, vi editor is used by default.

Example:

talosctl -n <IP> edit machineconfig

Configuration can be edited for multiple nodes if multiple IP addresses are specified:

talosctl -n <IP1>,<IP2>,... edit machineconfig

Applying machine configuration change immediately (without a reboot):

talosctl -n <IP> edit machineconfig --mode=no-reboot

talosctl patch machineconfig

Command talosctl patch works similar to talosctl edit command - it loads current machine configuration, but instead of launching configured editor it applies a set of JSON patches to the configuration and writes the result back to the node.

Example, updating kubelet version (in auto mode):

$ talosctl -n <IP> patch machineconfig -p '[{"op": "replace", "path": "/machine/kubelet/image", "value": "ghcr.io/siderolabs/kubelet:v1.31.1"}]'
patched mc at the node <IP>

Updating kube-apiserver version in immediate mode (without a reboot):

$ talosctl -n <IP> patch machineconfig --mode=no-reboot -p '[{"op": "replace", "path": "/cluster/apiServer/image", "value": "registry.k8s.io/kube-apiserver:v1.31.1"}]'
patched mc at the node <IP>

A patch might be applied to multiple nodes when multiple IPs are specified:

talosctl -n <IP1>,<IP2>,... patch machineconfig -p '[{...}]'

Patches can also be sourced from files using @file syntax:

talosctl -n <IP> patch machineconfig -p @kubelet-patch.json -p @manifest-patch.json

It might be easier to store patches in YAML format vs. the default JSON format. Talos can detect file format automatically:

# kubelet-patch.yaml
- op: replace
  path: /machine/kubelet/image
  value: ghcr.io/siderolabs/kubelet:v1.31.1
talosctl -n <IP> patch machineconfig -p @kubelet-patch.yaml

Recovering from Node Boot Failures

If a Talos node fails to boot because of wrong configuration (for example, control plane endpoint is incorrect), configuration can be updated to fix the issue.

7 - Logging

Dealing with Talos Linux logs.

Viewing logs

Kernel messages can be retrieved with talosctl dmesg command:

$ talosctl -n 172.20.1.2 dmesg

172.20.1.2: kern:    info: [2021-11-10T10:09:37.662764956Z]: Command line: init_on_alloc=1 slab_nomerge pti=on consoleblank=0 nvme_core.io_timeout=4294967295 printk.devkmsg=on ima_template=ima-ng ima_appraise=fix ima_hash=sha512 reboot=k panic=1 talos.shutdown=halt talos.platform=metal talos.config=http://172.20.1.1:40101/config.yaml
[...]

Service logs can be retrieved with talosctl logs command:

$ talosctl -n 172.20.1.2 services

NODE         SERVICE      STATE     HEALTH   LAST CHANGE   LAST EVENT
172.20.1.2   apid         Running   OK       19m27s ago    Health check successful
172.20.1.2   containerd   Running   OK       19m29s ago    Health check successful
172.20.1.2   cri          Running   OK       19m27s ago    Health check successful
172.20.1.2   etcd         Running   OK       19m22s ago    Health check successful
172.20.1.2   kubelet      Running   OK       19m20s ago    Health check successful
172.20.1.2   machined     Running   ?        19m30s ago    Service started as goroutine
172.20.1.2   trustd       Running   OK       19m27s ago    Health check successful
172.20.1.2   udevd        Running   OK       19m28s ago    Health check successful

$ talosctl -n 172.20.1.2 logs machined

172.20.1.2: [talos] task setupLogger (1/1): done, 106.109µs
172.20.1.2: [talos] phase logger (1/7): done, 564.476µs
[...]

Container logs for Kubernetes pods can be retrieved with talosctl logs -k command:

$ talosctl -n 172.20.1.2 containers -k
NODE         NAMESPACE   ID                                                              IMAGE                                                         PID    STATUS
172.20.1.2   k8s.io      kube-system/kube-flannel-dk6d5                                  registry.k8s.io/pause:3.6                                     1329   SANDBOX_READY
172.20.1.2   k8s.io      └─ kube-system/kube-flannel-dk6d5:install-cni:f1d4cf68feb9      ghcr.io/siderolabs/install-cni:v0.7.0-alpha.0-1-g2bb2efc      0      CONTAINER_EXITED
172.20.1.2   k8s.io      └─ kube-system/kube-flannel-dk6d5:install-config:bc39fec3cbac   quay.io/coreos/flannel:v0.13.0                                0      CONTAINER_EXITED
172.20.1.2   k8s.io      └─ kube-system/kube-flannel-dk6d5:kube-flannel:5c3989353b98     quay.io/coreos/flannel:v0.13.0                                1610   CONTAINER_RUNNING
172.20.1.2   k8s.io      kube-system/kube-proxy-gfkqj                                    registry.k8s.io/pause:3.5                                     1311   SANDBOX_READY
172.20.1.2   k8s.io      └─ kube-system/kube-proxy-gfkqj:kube-proxy:ad5e8ddc7e7f         registry.k8s.io/kube-proxy:v1.31.1                            1379   CONTAINER_RUNNING

$ talosctl -n 172.20.1.2 logs -k kube-system/kube-proxy-gfkqj:kube-proxy:ad5e8ddc7e7f
172.20.1.2: 2021-11-30T19:13:20.567825192Z stderr F I1130 19:13:20.567737       1 server_others.go:138] "Detected node IP" address="172.20.0.3"
172.20.1.2: 2021-11-30T19:13:20.599684397Z stderr F I1130 19:13:20.599613       1 server_others.go:206] "Using iptables Proxier"
[...]

If some host workloads (e.g. system extensions) send syslog messages, they can be retrieved with talosctl logs syslogd command.

Sending logs

Service logs

You can enable logs sendings in machine configuration:

machine:
  logging:
    destinations:
      - endpoint: "udp://127.0.0.1:12345/"
        format: "json_lines"
      - endpoint: "tcp://host:5044/"
        format: "json_lines"

Several destinations can be specified. Supported protocols are UDP and TCP. The only currently supported format is json_lines:

{
  "msg": "[talos] apply config request: immediate true, on reboot false",
  "talos-level": "info",
  "talos-service": "machined",
  "talos-time": "2021-11-10T10:48:49.294858021Z"
}

Messages are newline-separated when sent over TCP. Over UDP messages are sent with one message per packet. msg, talos-level, talos-service, and talos-time fields are always present; there may be additional fields.

Every message sent can be enhanced with additional fields by using the extraTags field in the machine configuration:

machine:
  logging:
    destinations:
      - endpoint: "udp://127.0.0.1:12345/"
        format: "json_lines"
        extraTags:
          server: s03-rack07

The specified extraTags are added to every message sent to the destination verbatim.

Kernel logs

Kernel log delivery can be enabled with the talos.logging.kernel kernel command line argument, which can be specified in the .machine.installer.extraKernelArgs:

machine:
  install:
    extraKernelArgs:
      - talos.logging.kernel=tcp://host:5044/

Also kernel logs delivery can be configured using the document in machine configuration:

apiVersion: v1alpha1
kind: KmsgLogConfig
name: remote-log
url: tcp://host:5044/

Kernel log destination is specified in the same way as service log endpoint. The only supported format is json_lines.

Sample message:

{
  "clock":6252819, // time relative to the kernel boot time
  "facility":"user",
  "msg":"[talos] task startAllServices (1/1): waiting for 6 services\n",
  "priority":"warning",
  "seq":711,
  "talos-level":"warn", // Talos-translated `priority` into common logging level
  "talos-time":"2021-11-26T16:53:21.3258698Z" // Talos-translated `clock` using current time
}

extraKernelArgs in the machine configuration are only applied on Talos upgrades, not just by applying the config. (Upgrading to the same version is fine).

Filebeat example

To forward logs to other Log collection services, one way to do this is sending them to a Filebeat running in the cluster itself (in the host network), which takes care of forwarding it to other endpoints (and the necessary transformations).

If Elastic Cloud on Kubernetes is being used, the following Beat (custom resource) configuration might be helpful:

apiVersion: beat.k8s.elastic.co/v1beta1
kind: Beat
metadata:
  name: talos
spec:
  type: filebeat
  version: 7.15.1
  elasticsearchRef:
    name: talos
  config:
    filebeat.inputs:
      - type: "udp"
        host: "127.0.0.1:12345"
        processors:
          - decode_json_fields:
              fields: ["message"]
              target: ""
          - timestamp:
              field: "talos-time"
              layouts:
                - "2006-01-02T15:04:05.999999999Z07:00"
          - drop_fields:
              fields: ["message", "talos-time"]
          - rename:
              fields:
                - from: "msg"
                  to: "message"

  daemonSet:
    updateStrategy:
      rollingUpdate:
        maxUnavailable: 100%
    podTemplate:
      spec:
        dnsPolicy: ClusterFirstWithHostNet
        hostNetwork: true
        securityContext:
          runAsUser: 0
        containers:
          - name: filebeat
            ports:
              - protocol: UDP
                containerPort: 12345
                hostPort: 12345

The input configuration ensures that messages and timestamps are extracted properly. Refer to the Filebeat documentation on how to forward logs to other outputs.

Also note the hostNetwork: true in the daemonSet configuration.

This ensures filebeat uses the host network, and listens on 127.0.0.1:12345 (UDP) on every machine, which can then be specified as a logging endpoint in the machine configuration.

Fluent-bit example

First, we’ll create a value file for the fluentd-bit Helm chart.

# fluentd-bit.yaml

podAnnotations:
  fluentbit.io/exclude: 'true'

extraPorts:
  - port: 12345
    containerPort: 12345
    protocol: TCP
    name: talos

config:
  service: |
    [SERVICE]
      Flush         5
      Daemon        Off
      Log_Level     warn
      Parsers_File  custom_parsers.conf    

  inputs: |
    [INPUT]
      Name          tcp
      Listen        0.0.0.0
      Port          12345
      Format        json
      Tag           talos.*

    [INPUT]
      Name          tail
      Alias         kubernetes
      Path          /var/log/containers/*.log
      Parser        containerd
      Tag           kubernetes.*

    [INPUT]
      Name          tail
      Alias         audit
      Path          /var/log/audit/kube/*.log
      Parser        audit
      Tag           audit.*    

  filters: |
    [FILTER]
      Name                kubernetes
      Alias               kubernetes
      Match               kubernetes.*
      Kube_Tag_Prefix     kubernetes.var.log.containers.
      Use_Kubelet         Off
      Merge_Log           On
      Merge_Log_Trim      On
      Keep_Log            Off
      K8S-Logging.Parser  Off
      K8S-Logging.Exclude On
      Annotations         Off
      Labels              On

    [FILTER]
      Name          modify
      Match         kubernetes.*
      Add           source kubernetes
      Remove        logtag    

  customParsers: |
    [PARSER]
      Name          audit
      Format        json
      Time_Key      requestReceivedTimestamp
      Time_Format   %Y-%m-%dT%H:%M:%S.%L%z

    [PARSER]
      Name          containerd
      Format        regex
      Regex         ^(?<time>[^ ]+) (?<stream>stdout|stderr) (?<logtag>[^ ]*) (?<log>.*)$
      Time_Key      time
      Time_Format   %Y-%m-%dT%H:%M:%S.%L%z    

  outputs: |
    [OUTPUT]
      Name    stdout
      Alias   stdout
      Match   *
      Format  json_lines    

  # If you wish to ship directly to Loki from Fluentbit,
  # Uncomment the following output, updating the Host with your Loki DNS/IP info as necessary.
  # [OUTPUT]
  # Name loki
  # Match *
  # Host loki.loki.svc
  # Port 3100
  # Labels job=fluentbit
  # Auto_Kubernetes_Labels on

daemonSetVolumes:
  - name: varlog
    hostPath:
      path: /var/log

daemonSetVolumeMounts:
  - name: varlog
    mountPath: /var/log

tolerations:
  - operator: Exists
    effect: NoSchedule

Next, we will add the helm repo for FluentBit, and deploy it to the cluster.

helm repo add fluent https://fluent.github.io/helm-charts
helm upgrade -i --namespace=kube-system -f fluentd-bit.yaml fluent-bit fluent/fluent-bit

Now we need to find the service IP.

$ kubectl -n kube-system get svc -l app.kubernetes.io/name=fluent-bit

NAME         TYPE        CLUSTER-IP     EXTERNAL-IP   PORT(S)             AGE
fluent-bit   ClusterIP   10.200.0.138   <none>        2020/TCP,5170/TCP   108m

Finally, we will change talos log destination with the command talosctl edit mc.

machine:
  logging:
    destinations:
      - endpoint: "tcp://10.200.0.138:5170"
        format: "json_lines"

This example configuration was well tested with Cilium CNI, and it should work with iptables/ipvs based CNI plugins too.

Vector example

Vector is a lightweight observability pipeline ideal for a Kubernetes environment. It can ingest (source) logs from multiple sources, perform remapping on the logs (transform), and forward the resulting pipeline to multiple destinations (sinks). As it is an end to end platform, it can be run as a single-deployment ‘aggregator’ as well as a replicaSet of ‘Agents’ that run on each node.

As Talos can be set as above to send logs to a destination, we can run Vector as an Aggregator, and forward both kernel and service to a UDP socket in-cluster.

Below is an excerpt of a source/sink setup for Talos, with a ‘sink’ destination of an in-cluster Grafana Loki log aggregation service. As Loki can create labels from the log input, we have set up the Loki sink to create labels based on the host IP, service and facility of the inbound logs.

Note that a method of exposing the Vector service will be required which may vary depending on your setup - a LoadBalancer is a good option.

role: "Stateless-Aggregator"

# Sources
sources:
  talos_kernel_logs:
    address: 0.0.0.0:6050
    type: socket
    mode: udp
    max_length: 102400
    decoding:
      codec: json
    host_key: __host

  talos_service_logs:
    address: 0.0.0.0:6051
    type: socket
    mode: udp
    max_length: 102400
    decoding:
      codec: json
    host_key: __host

# Sinks
sinks:
  talos_kernel:
    type: loki
    inputs:
      - talos_kernel_logs_xform
    endpoint: http://loki.system-monitoring:3100
    encoding:
      codec: json
      except_fields:
        - __host
    batch:
      max_bytes: 1048576
    out_of_order_action: rewrite_timestamp
    labels:
      hostname: >-
                {{`{{ __host }}`}}
      facility: >-
                {{`{{ facility }}`}}

  talos_service:
    type: loki
    inputs:
      - talos_service_logs_xform
    endpoint: http://loki.system-monitoring:3100
    encoding:
      codec: json
      except_fields:
        - __host
    batch:
      max_bytes: 400000
    out_of_order_action: rewrite_timestamp
    labels:
      hostname: >-
                {{`{{ __host }}`}}
      service: >-
                {{`{{ "talos-service" }}`}}

8 - NVIDIA Fabric Manager

In this guide we’ll follow the procedure to enable NVIDIA Fabric Manager.

NVIDIA GPUs that have nvlink support (for eg: A100) will need the nvidia-fabricmanager system extension also enabled in addition to the NVIDIA drivers. For more information on Fabric Manager refer https://docs.nvidia.com/datacenter/tesla/fabric-manager-user-guide/index.html

The published versions of the NVIDIA fabricmanager system extensions is available here

The nvidia-fabricmanager extension version has to match with the NVIDIA driver version in use.

Enabling the NVIDIA fabricmanager system extension

Create the boot assets or a custom installer and perform a machine upgrade which include the following system extensions:

ghcr.io/siderolabs/nvidia-open-gpu-kernel-modules-lts:535.183.06-v1.8.0
ghcr.io/siderolabs/nvidia-container-toolkit-lts:535.183.06-v1.16.1
ghcr.io/siderolabs/nvidia-fabricmanager:535.183.06

Patch the machine configuration to load the required modules:

machine:
  kernel:
    modules:
      - name: nvidia
      - name: nvidia_uvm
      - name: nvidia_drm
      - name: nvidia_modeset
  sysctls:
    net.core.bpf_jit_harden: 1

9 - NVIDIA GPU (OSS drivers)

In this guide we’ll follow the procedure to support NVIDIA GPU using OSS drivers on Talos.

Enabling NVIDIA GPU support on Talos is bound by NVIDIA EULA. The Talos published NVIDIA OSS drivers are bound to a specific Talos release. The extensions versions also needs to be updated when upgrading Talos.

We will be using the following NVIDIA OSS system extensions:

  • nvidia-open-gpu-kernel-modules
  • nvidia-container-toolkit

Create the boot assets which includes the system extensions mentioned above (or create a custom installer and perform a machine upgrade if Talos is already installed).

Make sure the driver version matches for both the nvidia-open-gpu-kernel-modules and nvidia-container-toolkit extensions. The nvidia-open-gpu-kernel-modules extension is versioned as <nvidia-driver-version>-<talos-release-version> and the nvidia-container-toolkit extension is versioned as <nvidia-driver-version>-<nvidia-container-toolkit-version>.

Proprietary vs OSS Nvidia Driver Support

The NVIDIA Linux GPU Driver contains several kernel modules: nvidia.ko, nvidia-modeset.ko, nvidia-uvm.ko, nvidia-drm.ko, and nvidia-peermem.ko. Two “flavors” of these kernel modules are provided, and both are available for use within Talos:

The choice between Proprietary/OSS may be decided after referencing the Official NVIDIA announcement.

Enabling the NVIDIA OSS modules

Patch Talos machine configuration using the patch gpu-worker-patch.yaml:

machine:
  kernel:
    modules:
      - name: nvidia
      - name: nvidia_uvm
      - name: nvidia_drm
      - name: nvidia_modeset
  sysctls:
    net.core.bpf_jit_harden: 1

Now apply the patch to all Talos nodes in the cluster having NVIDIA GPU’s installed:

talosctl patch mc --patch @gpu-worker-patch.yaml

The NVIDIA modules should be loaded and the system extension should be installed.

This can be confirmed by running:

talosctl read /proc/modules

which should produce an output similar to below:

nvidia_uvm 1146880 - - Live 0xffffffffc2733000 (PO)
nvidia_drm 69632 - - Live 0xffffffffc2721000 (PO)
nvidia_modeset 1142784 - - Live 0xffffffffc25ea000 (PO)
nvidia 39047168 - - Live 0xffffffffc00ac000 (PO)
talosctl get extensions

which should produce an output similar to below:

NODE           NAMESPACE   TYPE              ID                                                                           VERSION   NAME                             VERSION
172.31.41.27   runtime     ExtensionStatus   000.ghcr.io-siderolabs-nvidia-container-toolkit-515.65.01-v1.10.0            1         nvidia-container-toolkit         515.65.01-v1.10.0
172.31.41.27   runtime     ExtensionStatus   000.ghcr.io-siderolabs-nvidia-open-gpu-kernel-modules-515.65.01-v1.2.0       1         nvidia-open-gpu-kernel-modules   515.65.01-v1.2.0
talosctl read /proc/driver/nvidia/version

which should produce an output similar to below:

NVRM version: NVIDIA UNIX x86_64 Kernel Module  515.65.01  Wed Mar 16 11:24:05 UTC 2022
GCC version:  gcc version 12.2.0 (GCC)

Deploying NVIDIA device plugin

First we need to create the RuntimeClass

Apply the following manifest to create a runtime class that uses the extension:

---
apiVersion: node.k8s.io/v1
kind: RuntimeClass
metadata:
  name: nvidia
handler: nvidia

Install the NVIDIA device plugin:

helm repo add nvdp https://nvidia.github.io/k8s-device-plugin
helm repo update
helm install nvidia-device-plugin nvdp/nvidia-device-plugin --version=0.13.0 --set=runtimeClassName=nvidia

(Optional) Setting the default runtime class as nvidia

Do note that this will set the default runtime class to nvidia for all pods scheduled on the node.

Create a patch yaml nvidia-default-runtimeclass.yaml to update the machine config similar to below:

- op: add
  path: /machine/files
  value:
    - content: |
        [plugins]
          [plugins."io.containerd.cri.v1.runtime"]
            [plugins."io.containerd.cri.v1.runtime".containerd]
              default_runtime_name = "nvidia"        
      path: /etc/cri/conf.d/20-customization.part
      op: create

Now apply the patch to all Talos nodes in the cluster having NVIDIA GPU’s installed:

talosctl patch mc --patch @nvidia-default-runtimeclass.yaml

Testing the runtime class

Note the spec.runtimeClassName being explicitly set to nvidia in the pod spec.

Run the following command to test the runtime class:

kubectl run \
  nvidia-test \
  --restart=Never \
  -ti --rm \
  --image nvcr.io/nvidia/cuda:12.5.0-base-ubuntu22.04 \
  --overrides '{"spec": {"runtimeClassName": "nvidia"}}' \
  nvidia-smi

10 - NVIDIA GPU (Proprietary drivers)

In this guide we’ll follow the procedure to support NVIDIA GPU using proprietary drivers on Talos.

Enabling NVIDIA GPU support on Talos is bound by NVIDIA EULA. The Talos published NVIDIA drivers are bound to a specific Talos release. The extensions versions also needs to be updated when upgrading Talos.

We will be using the following NVIDIA system extensions:

  • nonfree-kmod-nvidia
  • nvidia-container-toolkit

To build a NVIDIA driver version not published by SideroLabs follow the instructions here

Create the boot assets which includes the system extensions mentioned above (or create a custom installer and perform a machine upgrade if Talos is already installed).

Make sure the driver version matches for both the nonfree-kmod-nvidia and nvidia-container-toolkit extensions. The nonfree-kmod-nvidia extension is versioned as <nvidia-driver-version>-<talos-release-version> and the nvidia-container-toolkit extension is versioned as <nvidia-driver-version>-<nvidia-container-toolkit-version>.

Proprietary vs OSS Nvidia Driver Support

The NVIDIA Linux GPU Driver contains several kernel modules: nvidia.ko, nvidia-modeset.ko, nvidia-uvm.ko, nvidia-drm.ko, and nvidia-peermem.ko. Two “flavors” of these kernel modules are provided, and both are available for use within Talos:

The choice between Proprietary/OSS may be decided after referencing the Official NVIDIA announcement.

Enabling the NVIDIA modules and the system extension

Patch Talos machine configuration using the patch gpu-worker-patch.yaml:

machine:
  kernel:
    modules:
      - name: nvidia
      - name: nvidia_uvm
      - name: nvidia_drm
      - name: nvidia_modeset
  sysctls:
    net.core.bpf_jit_harden: 1

Now apply the patch to all Talos nodes in the cluster having NVIDIA GPU’s installed:

talosctl patch mc --patch @gpu-worker-patch.yaml

The NVIDIA modules should be loaded and the system extension should be installed.

This can be confirmed by running:

talosctl read /proc/modules

which should produce an output similar to below:

nvidia_uvm 1146880 - - Live 0xffffffffc2733000 (PO)
nvidia_drm 69632 - - Live 0xffffffffc2721000 (PO)
nvidia_modeset 1142784 - - Live 0xffffffffc25ea000 (PO)
nvidia 39047168 - - Live 0xffffffffc00ac000 (PO)
talosctl get extensions

which should produce an output similar to below:

NODE           NAMESPACE   TYPE              ID                                                                 VERSION   NAME                       VERSION
172.31.41.27   runtime     ExtensionStatus   000.ghcr.io-frezbo-nvidia-container-toolkit-510.60.02-v1.9.0       1         nvidia-container-toolkit   510.60.02-v1.9.0
talosctl read /proc/driver/nvidia/version

which should produce an output similar to below:

NVRM version: NVIDIA UNIX x86_64 Kernel Module  510.60.02  Wed Mar 16 11:24:05 UTC 2022
GCC version:  gcc version 11.2.0 (GCC)

Deploying NVIDIA device plugin

First we need to create the RuntimeClass

Apply the following manifest to create a runtime class that uses the extension:

---
apiVersion: node.k8s.io/v1
kind: RuntimeClass
metadata:
  name: nvidia
handler: nvidia

Install the NVIDIA device plugin:

helm repo add nvdp https://nvidia.github.io/k8s-device-plugin
helm repo update
helm install nvidia-device-plugin nvdp/nvidia-device-plugin --version=0.13.0 --set=runtimeClassName=nvidia

(Optional) Setting the default runtime class as nvidia

Do note that this will set the default runtime class to nvidia for all pods scheduled on the node.

Create a patch yaml nvidia-default-runtimeclass.yaml to update the machine config similar to below:

- op: add
  path: /machine/files
  value:
    - content: |
        [plugins]
          [plugins."io.containerd.cri.v1.runtime"]
            [plugins."io.containerd.cri.v1.runtime".containerd]
              default_runtime_name = "nvidia"        
      path: /etc/cri/conf.d/20-customization.part
      op: create

Now apply the patch to all Talos nodes in the cluster having NVIDIA GPU’s installed:

talosctl patch mc --patch @nvidia-default-runtimeclass.yaml

Testing the runtime class

Note the spec.runtimeClassName being explicitly set to nvidia in the pod spec.

Run the following command to test the runtime class:

kubectl run \
  nvidia-test \
  --restart=Never \
  -ti --rm \
  --image nvcr.io/nvidia/cuda:12.5.0-base-ubuntu22.04 \
  --overrides '{"spec": {"runtimeClassName": "nvidia"}}' \
  nvidia-smi

11 - Performance Tuning

In this guide, we’ll describe various performance tuning knobs available.

Talos Linux tries to strike a balance between performance and security/efficiency. However, there are some performance tuning knobs available to adjust the system to your needs. With any performance tuning, it’s essential to measure the impact of the changes and ensure they don’t introduce security vulnerabilities.

Note: Most of the suggestions below apply to bare metal machines, but some of them might be useful for VMs as well.

If you find more performance tuning knobs, please let us know by editing this document.

Kernel Parameters

Talos Linux kernel parameters can be adjusted in the following ways:

  • temporary, one-time adjustments can be done via console access, and editing the kernel command line in the bootloader (doesn’t work for Secure Boot enabled systems)
  • on initial install (when booting off ISO/PXE), .machine.install.extraKernelArgs can be used to set kernel parameters
  • after the initial install (or when booting off a disk image), .machine.install.extraKernelArgs changes require a no-op upgrade (e.g. to the same version of Talos) to take effect

CPU Scaling

Talos Linux uses the schedutil CPU scaling governor by default, for maximum performance, you can switch to the performance governor:

cpufreq.default_governor=performance

Processor Sleep States

Modern processors support various sleep states to save power, but they might introduce latency when transitioning back to the active state.

AMD

For maximum performance (and lower latency), use active mode of the amd-pstate driver:

amd_pstate=active

Intel

For maximum performance (and lower latency), disable the intel_idle driver:

intel_idle.max_cstate=0

Hardware Vulnerabilities

Modern processors have various security vulnerabilities that require software/microcode mitigations. These mitigations might have a performance impact, and some of them can be disabled if you are willing to take the risk.

First of all, ensure that Talos system extensions amd-ucode and intel-ucode are installed (and using latest version of Talos Linux). Linux kernel will load the microcode updates on early boot, and for some processors, it might reduce the performance impact of the mitigations. The availability of microcode updates depends on the processor model.

The kernel command line argument mitigations can be used to disable all mitigations at once (not recommended from security point of view):

mitigations=off

There is also a way to disable specific mitigations, see Kernel documentation for more details.

I/O

For Talos Linux before version 1.8.2, the I/O performance can be improved by setting iommu.strict=0, for later versions this is a default setting.

Performance can be further improved at some cost of security by bypassing the I/O memory management unit (IOMMU) for DMA:

iommu.passthrough=1

12 - Pull Through Image Cache

How to set up local transparent container images caches.

In this guide we will create a set of local caching Docker registry proxies to minimize local cluster startup time.

When running Talos locally, pulling images from container registries might take a significant amount of time. We spin up local caching pass-through registries to cache images and configure a local Talos cluster to use those proxies. A similar approach might be used to run Talos in production in air-gapped environments. It can be also used to verify that all the images are available in local registries.

Video Walkthrough

To see a live demo of this writeup, see the video below:

Requirements

The follow are requirements for creating the set of caching proxies:

  • Docker 18.03 or greater
  • Local cluster requirements for either docker or QEMU.

Launch the Caching Docker Registry Proxies

Talos pulls from docker.io, registry.k8s.io, gcr.io, and ghcr.io by default. If your configuration is different, you might need to modify the commands below:

docker run -d -p 5000:5000 \
    -e REGISTRY_PROXY_REMOTEURL=https://registry-1.docker.io \
    --restart always \
    --name registry-docker.io registry:2

docker run -d -p 5001:5000 \
    -e REGISTRY_PROXY_REMOTEURL=https://registry.k8s.io \
    --restart always \
    --name registry-registry.k8s.io registry:2

docker run -d -p 5003:5000 \
    -e REGISTRY_PROXY_REMOTEURL=https://gcr.io \
    --restart always \
    --name registry-gcr.io registry:2

docker run -d -p 5004:5000 \
    -e REGISTRY_PROXY_REMOTEURL=https://ghcr.io \
    --restart always \
    --name registry-ghcr.io registry:2

Note: Proxies are started as docker containers, and they’re automatically configured to start with Docker daemon.

As a registry container can only handle a single upstream Docker registry, we launch a container per upstream, each on its own host port (5000, 5001, 5002, 5003 and 5004).

Using Caching Registries with QEMU Local Cluster

With a QEMU local cluster, a bridge interface is created on the host. As registry containers expose their ports on the host, we can use bridge IP to direct proxy requests.

sudo talosctl cluster create --provisioner qemu \
    --registry-mirror docker.io=http://10.5.0.1:5000 \
    --registry-mirror registry.k8s.io=http://10.5.0.1:5001 \
    --registry-mirror gcr.io=http://10.5.0.1:5003 \
    --registry-mirror ghcr.io=http://10.5.0.1:5004

The Talos local cluster should now start pulling via caching registries. This can be verified via registry logs, e.g. docker logs -f registry-docker.io. The first time cluster boots, images are pulled and cached, so next cluster boot should be much faster.

Note: 10.5.0.1 is a bridge IP with default network (10.5.0.0/24), if using custom --cidr, value should be adjusted accordingly.

Using Caching Registries with docker Local Cluster

With a docker local cluster we can use docker bridge IP, default value for that IP is 172.17.0.1. On Linux, the docker bridge address can be inspected with ip addr show docker0.

talosctl cluster create --provisioner docker \
    --registry-mirror docker.io=http://172.17.0.1:5000 \
    --registry-mirror registry.k8s.io=http://172.17.0.1:5001 \
    --registry-mirror gcr.io=http://172.17.0.1:5003 \
    --registry-mirror ghcr.io=http://172.17.0.1:5004

Machine Configuration

The caching registries can be configured via machine configuration patch, equivalent to the command line flags above:

machine:
  registries:
    mirrors:
      docker.io:
        endpoints:
          - http://10.5.0.1:5000
      gcr.io:
        endpoints:
          - http://10.5.0.1:5003
      ghcr.io:
        endpoints:
          - http://10.5.0.1:5004
      registry.k8s.io:
        endpoints:
          - http://10.5.0.1:5001

Cleaning Up

To cleanup, run:

docker rm -f registry-docker.io
docker rm -f registry-registry.k8s.io
docker rm -f registry-gcr.io
docker rm -f registry-ghcr.io

Note: Removing docker registry containers also removes the image cache. So if you plan to use caching registries, keep the containers running.

Using Harbor as a Caching Registry

Harbor is an open source container registry that can be used as a caching proxy. Harbor supports configuring multiple upstream registries, so it can be used to cache multiple registries at once behind a single endpoint.

Harbor Endpoints

Harbor Projects

As Harbor puts a registry name in the pull image path, we need to set overridePath: true to prevent Talos and containerd from appending /v2 to the path.

machine:
  registries:
    mirrors:
      docker.io:
        endpoints:
          - http://harbor/v2/proxy-docker.io
        overridePath: true
      ghcr.io:
        endpoints:
          - http://harbor/v2/proxy-ghcr.io
        overridePath: true
      gcr.io:
        endpoints:
          - http://harbor/v2/proxy-gcr.io
        overridePath: true
      registry.k8s.io:
        endpoints:
          - http://harbor/v2/proxy-registry.k8s.io
        overridePath: true

The Harbor external endpoint (http://harbor in this example) can be configured with authentication or custom TLS:

machine:
  registries:
    config:
      harbor:
        auth:
          username: admin
          password: password

13 - Role-based access control (RBAC)

Set up RBAC on the Talos Linux API.

Talos v0.11 introduced initial support for role-based access control (RBAC). This guide will explain what that is and how to enable it without losing access to the cluster.

RBAC in Talos

Talos uses certificates to authorize users. The certificate subject’s organization field is used to encode user roles. There is a set of predefined roles that allow access to different API methods:

  • os:admin grants access to all methods;
  • os:operator grants everything os:reader role does, plus additional methods: rebooting, shutting down, etcd backup, etcd alarm management, and so on;
  • os:reader grants access to “safe” methods (for example, that includes the ability to list files, but does not include the ability to read files content);
  • os:etcd:backup grants access to /machine.MachineService/EtcdSnapshot method.

Roles in the current talosconfig can be checked with the following command:

$ talosctl config info

[...]
Roles:               os:admin
[...]

RBAC is enabled by default in new clusters created with talosctl v0.11+ and disabled otherwise.

Enabling RBAC

First, both the Talos cluster and talosctl tool should be upgraded. Then the talosctl config new command should be used to generate a new client configuration with the os:admin role. Additional configurations and certificates for different roles can be generated by passing --roles flag:

talosctl config new --roles=os:reader reader

That command will create a new client configuration file reader with a new certificate with os:reader role.

After that, RBAC should be enabled in the machine configuration:

machine:
  features:
    rbac: true

14 - System Extensions

Customizing the Talos Linux immutable root file system.

System extensions allow extending the Talos root filesystem, which enables a variety of features, such as including custom container runtimes, loading additional firmware, etc.

System extensions are only activated during the installation or upgrade of Talos Linux. With system extensions installed, the Talos root filesystem is still immutable and read-only.

Installing System Extensions

Note: the way to install system extensions in the .machine.install section of the machine configuration is now deprecated.

Starting with Talos v1.5.0, Talos supports generation of boot media with system extensions included, this removes the need to rebuild the initramfs.xz on the machine itself during the installation or upgrade.

There are two kinds of boot assets that Talos can generate:

  • initial boot assets (ISO, PXE, etc.) that are used to boot the machine
  • disk images that have Talos pre-installed
  • installer container images that can be used to install or upgrade Talos on a machine (installation happens when booted from ISO or PXE)

Depending on the nature of the system extension (e.g. network device driver or containerd plugin), it may be necessary to include the extension in both initial boot assets and disk images/installer, or just the installer.

The process of generating boot assets with extensions included is described in the boot assets guide.

Example: Booting from an ISO

Let’s assume NVIDIA extension is required on a bare metal machine which is going to be booted from an ISO. As NVIDIA extension is not required for the initial boot and install step, it is sufficient to include the extension in the installer image only.

  1. Use a generic Talos ISO to boot the machine.
  2. Prepare a custom installer container image with NVIDIA extension included, push the image to a registry.
  3. Ensure that machine configuration field .machine.install.image points to the custom installer image.
  4. Boot the machine using the ISO, apply the machine configuration.
  5. Talos pulls a custom installer image from the registry (containing NVIDIA extension), installs Talos on the machine, and reboots.

When it’s time to upgrade Talos, generate a custom installer container for a new version of Talos, push it to a registry, and perform upgrade pointing to the custom installer image.

Example: Disk Image

Let’s assume NVIDIA extension is required on AWS VM.

  1. Prepare an AWS disk image with NVIDIA extension included.
  2. Upload the image to AWS, register it as an AMI.
  3. Use the AMI to launch a VM.
  4. Talos boots with NVIDIA extension included.

When it’s time to upgrade Talos, either repeat steps 1-4 to replace the VM with a new AMI, or like in the previous example, generate a custom installer and use it to upgrade Talos in-place.

Authoring System Extensions

A Talos system extension is a container image with the specific folder structure. System extensions can be built and managed using any tool that produces container images, e.g. docker build.

Sidero Labs maintains a repository of system extensions.

Resource Definitions

Use talosctl get extensions to get a list of system extensions:

$ talosctl get extensions
NODE         NAMESPACE   TYPE              ID                                              VERSION   NAME          VERSION
172.20.0.2   runtime     ExtensionStatus   000.ghcr.io-talos-systems-gvisor-54b831d        1         gvisor        20220117.0-v1.0.0
172.20.0.2   runtime     ExtensionStatus   001.ghcr.io-talos-systems-intel-ucode-54b831d   1         intel-ucode   microcode-20210608-v1.0.0

Use YAML or JSON format to see additional details about the extension:

$ talosctl -n 172.20.0.2 get extensions 001.ghcr.io-talos-systems-intel-ucode-54b831d -o yaml
node: 172.20.0.2
metadata:
    namespace: runtime
    type: ExtensionStatuses.runtime.talos.dev
    id: 001.ghcr.io-talos-systems-intel-ucode-54b831d
    version: 1
    owner: runtime.ExtensionStatusController
    phase: running
    created: 2022-02-10T18:25:04Z
    updated: 2022-02-10T18:25:04Z
spec:
    image: 001.ghcr.io-talos-systems-intel-ucode-54b831d.sqsh
    metadata:
        name: intel-ucode
        version: microcode-20210608-v1.0.0
        author: Spencer Smith
        description: |
            This system extension provides Intel microcode binaries.
        compatibility:
            talos:
                version: '>= v1.0.0'

Example: gVisor

See readme of the gVisor extension.

15 - Time Synchronization

Configuring time synchronization.

Talos Linux itself does not require time to be synchronized across the cluster, but as Talos Linux and Kubernetes components issue certificates with expiration dates, it is recommended to have time synchronized across the cluster. Some workloads (e.g. Ceph) might require to be in sync across the machines in the cluster due to the design of the application.

Talos Linux tries to launch API even if the time is not sync, and if time jumps as a result of NTP sync, the API certificates will be rotated automatically. Some components like kubelet and etcd wait for the time to be in sync before starting, as they don’t support graceful certificate rotation.

By default, Talos Linux uses time.cloudflare.com as the NTP server, but it can be overridden in the machine configuration, or provided via DHCP, kernel args, platform sources, etc. Talos Linux implements SNTP protocol to sync time with the NTP server.

Observing Status

Current time sync status can be observed with:

$ talosctl get timestatus
NODE         NAMESPACE   TYPE         ID     VERSION   SYNCED
172.20.0.2   runtime     TimeStatus   node   2         true

The list of servers Talos Linux is syncing with can be observed with:

$ talosctl get timeservers
NODE         NAMESPACE   TYPE               ID            VERSION   TIMESERVERS
172.20.0.2   network     TimeServerStatus   timeservers   1         ["time.cloudflare.com"]

More detailed logs about the time sync process can be queried with:

$ talosctl logs controller-runtime | grep -i time.Sync
172.20.0.2: 2024-04-17T18:32:16.690Z DEBUG NTP response {"component": "controller-runtime", "controller": "time.SyncController", "clock_offset": "37.060204ms", "rtt": "3.044816ms", "leap": 0, "stratum": 3, "precision": "29ns", "root_delay": "70.617676ms", "root_dispersion": "259.399µs", "root_distance": "37.090645ms"}
172.20.0.2: 2024-04-17T18:32:16.690Z DEBUG sample stats {"component": "controller-runtime", "controller": "time.SyncController", "jitter": "150.196588ms", "poll_interval": "34m8s", "spike": false}
172.20.0.2: 2024-04-17T18:32:16.690Z DEBUG adjusting time (slew) by 37.060204ms via 162.159.200.1, state TIME_OK, status STA_PLL | STA_NANO {"component": "controller-runtime", "controller": "time.SyncController"}
172.20.0.2: 2024-04-17T18:32:16.690Z DEBUG adjtime state {"component": "controller-runtime", "controller": "time.SyncController", "constant": 7, "offset": "37.060203ms", "freq_offset": -1302069, "freq_offset_ppm": -19}

Using PTP Devices

When running in a VM on a hypervisor, instead of doing network time sync, Talos can sync the time to the hypervisor clock (if supported by the hypervisor).

To check if the PTP device is available:

$ talosctl ls /sys/class/ptp/
NODE         NAME
172.20.0.2   .
172.20.0.2   ptp0

Make sure that the PTP device is provided by the hypervisor, as some PTP devices don’t provide accurate time value without proper setup:

talosctl read /sys/class/ptp/ptp0/clock_name
KVM virtual PTP

To enable PTP sync, set the machine.time.servers to the PTP device name (e.g. /dev/ptp0):

machine:
  time:
    servers:
      - /dev/ptp0

After setting the PTP device, Talos will sync the time to the PTP device instead of using the NTP server:

172.20.0.2: 2024-04-17T19:11:48.817Z DEBUG adjusting time (slew) by 32.223689ms via /dev/ptp0, state TIME_OK, status STA_PLL | STA_NANO {"component": "controller-runtime", "controller": "time.SyncController"}

Additional Configuration

Talos NTP sync can be disabled with the following machine configuration patch:

machine:
  time:
    disabled: true

When time sync is disabled, Talos assumes that time is always in sync.

Time sync can be also configured on best-effort basis, where Talos will try to sync time for the specified period of time, but if it fails to do so, time will be configured to be in sync when the period expires:

machine:
  time:
    bootTimeout: 2m