Repository navigation
What's Changed
- Release v0.1.10 by @Handfish in #29
- Release 0.1.11: discovery-independent node enumeration + cluster-correct Kubernetes views by @Handfish in #31
Full Changelog: v0.1.10...v0.1.11
Release Notes
Reliability release: node enumeration no longer depends on the Talos discovery service, and Kubernetes views always target the launched cluster.
Bug Fixes
Worker nodes missing when cluster discovery is disabled (#25, #22)
0.1.10 fixed workers vanishing when --config broke context resolution, but they still disappeared whenever Talos cluster discovery returns no members. This is increasingly common: Sidero has disabled the Kubernetes discovery registry by default, leaving only the external discovery.talos.dev service registry — which hardened/airgapped clusters routinely turn off — so talosctl get members comes back empty as the norm, not the exception.
talos-pilot built its node roster from talosctl get members, and when that's empty it fell back to etcd membership — which only ever contains control-plane nodes — so workers dropped out (leaving the "cluster discovery returned no members" warning). It now falls back to the Kubernetes API, which knows every node (control plane and worker) independently of Talos discovery. The kubeconfig for that lookup is fetched directly from a control-plane node over the Talos API (certificate auth), so it always targets the correct cluster and never an unrelated one an ambient KUBECONFIG might point at. Displayed nodes still come from successful Talos queries, so unreachable/bogus IPs self-filter.
The same fallback is applied everywhere the app enumerates nodes independently of the main list:
- Cluster node list — workers appear again (and everything derived from it: rolling operations, per-node logs/processes/network/storage/operations).
- Lifecycle (versions / config drift) — the version-skew and config-drift table now covers all nodes, not just the endpoint the base client targets.
Reproduced on a discovery-disabled Docker cluster (1 control plane + 2 workers): before, both views showed the control plane only; after, all three nodes appear. Verified end-to-end + regression tests added.
Lifecycle "Config" and "Time Sync" columns were always blank
The Lifecycle node table showed - for every node's config version and time-sync status. Two causes: talosctl get machineconfig returns multiple YAML documents (the persistent and v1alpha1 configs), but the parser tried to decode the whole blob as a single document and always errored; and per-node time-sync info was only fetched for the endpoint. Both are fixed — machineconfig parsing now splits documents and prefers the canonical v1alpha1 config (tolerating a numeric version), and version + time-sync are queried per node — so the table is fully populated for control-plane and worker nodes alike. Regression tests added.
Kubernetes views could target the wrong cluster when KUBECONFIG was set
The workloads view and the drain/cordon operations build a Kubernetes client that preferred an ambient KUBECONFIG over the launched Talos context. If your shell's KUBECONFIG pointed at a different cluster than --context, read views silently showed the wrong cluster's data and — more seriously — drain/cordon could act on the wrong nodes.
These clients now pin to the launched cluster: the kubeconfig is fetched from a control-plane node over the Talos API, so it always identifies the cluster you asked for. An ambient KUBECONFIG is still honored when it names the same cluster (matched by cluster CA, so a VIP/load-balancer endpoint isn't mistaken for a different cluster) — preserving your configured, possibly more reachable endpoint — but is ignored when it names a different one. When it's ignored, a wrapped warning on the cluster view explains why, so nothing is silently overridden. Verified end-to-end against a cluster with a mismatched KUBECONFIG; regression tests added.
Install talos-pilot 0.1.11
Install prebuilt binaries via shell script
curl --proto '=https' --tlsv1.2 -LsSf https://github.com/Handfish/talos-pilot/releases/download/v0.1.11/talos-pilot-installer.sh | shInstall prebuilt binaries via powershell script
powershell -ExecutionPolicy Bypass -c "irm https://github.com/Handfish/talos-pilot/releases/download/v0.1.11/talos-pilot-installer.ps1 | iex"Install prebuilt binaries via Homebrew
brew install Handfish/tap/talos-pilotDownload talos-pilot 0.1.11
| File | Platform | Checksum |
|---|---|---|
| talos-pilot-aarch64-apple-darwin.tar.xz | Apple Silicon macOS | checksum |
| talos-pilot-x86_64-apple-darwin.tar.xz | Intel macOS | checksum |
| talos-pilot-x86_64-pc-windows-msvc.zip | x64 Windows | checksum |
| talos-pilot-aarch64-unknown-linux-gnu.tar.xz | ARM64 Linux | checksum |
| talos-pilot-x86_64-unknown-linux-gnu.tar.xz | x64 Linux | checksum |
| talos-pilot-aarch64-unknown-linux-musl.tar.xz | ARM64 MUSL Linux | checksum |
| talos-pilot-x86_64-unknown-linux-musl.tar.xz | x64 MUSL Linux | checksum |