What's New in Talos 0.10

Disaster Recovery

Talos now supports etcd snapshots and recovery from the snapshotted state. Periodic snapshots of etcd data can be taken with talosctl etcd snapshot command, and in case of catastrophic control plane failure etcd contents can be recovered from the latest snapshot with talosctl bootstrap --recover-from= command.

Time Synchronization

The timed service was replaced with a new time sync controller without any machine configuration changes. There should be no user-visible changes in the way new time synchronization process works, logs are now available via talosctl logs controller-runtime. Talos now prefers last successful time server (by IP address) on each sync attempt, which improves sync accuracy.

Single Board Computers

Talos added support for the Radxa Rock PI 4c board. u-boot version was updated to fix the boot and USB issues on Raspberry Pi 4 8GiB version.

Optimizations

Multiple optimizations were applied to reduce Talos initramfs size and memory footprint. As a result, we see a reduction of memory usage of around 100 MiB for the core Talos components which leaves more resources available for you workloads.

Install Disk Selector

Install section of the machine config now has diskSelector field that allows querying install disk using the list of qualifiers:

...
  install:
    diskSelector:
      size: >= 500GB
      model: WDC*
...

talosctl -n <IP> disks -i can be used to check allowed disk qualifiers when the node is running in the maintenance mode.

Inline Kubernetes Manifests

Kubernetes manifests can now be submitted in the machine configuration using the cluster.inlineManifests field, which works same way as cluster.extraManifests field, but manifest contents are passed inline in the machine configuration.

Updated Components

Linux: 5.10.19 -> 5.10.29

Kubernetes: 1.20.5 -> 1.21.0

Go: 1.15 -> 1.16