You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
SideroLink connects. A Machine is created immediately. No lasting Pending/Accept step.
• NodeUniqueTokenStatus.spec.state stays 0
• Link.spec.nodeuniquetoken is ""
• MachineStatus.hardware / network / talosversion stay null
• Omni never generates machine config
• Join-token warning: The machine node unique token status is not determined
• Generic Omni talosctl can get machinestatus / get disks (Talos is maintenance, mmcblk0 present)
• talosctl reboot|reset|apply-config fail: cluster "" endpoint not found
• omnictl machine install|shutdown fail: “neither in maintenance nor allocated”
• omnictl configure machine --reset-node-unique-token fails: no NodeUniqueTokens resource
• Admin user and Admin service account cannot GET NodeUniqueTokens or delete Machine (“only read access”)
Expected Behavior
On first SideroLink connect after a reimage (same board UUID, new SD):
After Accept, Omni creates a NodeUniqueToken, collects hardware/disks, and generates machine config.
If the token matches, the machine joins as a normal unallocated machine with inventory.
If the token mismatches, Omni rejects the join instead of creating a half-dead Machine.
While the node is in Talos maintenance, Omni treats it as maintenance so omnictl machine install / talosctl reboot through the generic Omni talosconfig work.
Steps To Reproduce
Omni SaaS (hobby), backend/omnictl 1.10.3. Existing cluster with Raspberry Pi control planes (board UUIDs like 30303031-3030-3030-6261-…).
Delete the control-plane MachineSet (etcd restore planned via ControlPlane.bootstrapSpec.snapshot).
Flash the Pi SD with the Omni Raspberry Pi disk image (same default join token) or stock Image Factory rpi_generic v1.13.8 + Omni Machine Join Config.
Boot the Pi. It reconnects on the same hardware UUID.
Observe: connected Machine, empty inventory, token state 0, no config. Optional: add it back to the control-plane set — it sits in Configuring forever.
Same result if the Machine/Link is deleted and the Pi is left to rejoin unallocated. PendingMachines may appear for a few seconds, then a Machine is created with no token.
Is there an existing issue for this?
Current Behavior
SideroLink connects. A Machine is created immediately. No lasting Pending/Accept step.
• NodeUniqueTokenStatus.spec.state stays 0
• Link.spec.nodeuniquetoken is ""
• MachineStatus.hardware / network / talosversion stay null
• Omni never generates machine config
• Join-token warning: The machine node unique token status is not determined
• Generic Omni talosctl can get machinestatus / get disks (Talos is maintenance, mmcblk0 present)
• talosctl reboot|reset|apply-config fail: cluster "" endpoint not found
• omnictl machine install|shutdown fail: “neither in maintenance nor allocated”
• omnictl configure machine --reset-node-unique-token fails: no NodeUniqueTokens resource
• Admin user and Admin service account cannot GET NodeUniqueTokens or delete Machine (“only read access”)
Expected Behavior
On first SideroLink connect after a reimage (same board UUID, new SD):
Steps To Reproduce
Same result if the Machine/Link is deleted and the Pi is left to rejoin unallocated. PendingMachines may appear for a few seconds, then a Machine is created with no token.
What browsers are you seeing the problem on?
Chrome
Anything else?
issue ongoing inside https://ap.omni.siderolabs.io/