Robot is offline

Where this comes from

This is the page <seyd-connect-error> links to for the diagnosis class robot-offline. The class is chosen by classify() in @seyd/web from the robot’s NAT report and the failure reason; see Reachability for how the report is built.

The robot id exists, but nothing is currently announced under it: the daemon is not running, has no internet access, or was refused by signalling.

What to do

1. Check that the agent is running on the robot (systemctl status seydd, or the process you started) and read its log.
2. Look for denied in the log. unknown-robot means the robot was never enrolled; redeem an enrolment token (see Enrolment and access). key-mismatch means the id is enrolled with a different key: another agent is using the id, or the credential file was replaced. Delete the robot in the console and enrol again.
3. After a cloud deploy, a robot can show offline for a while: its signalling WebSocket stays on the draining old revision until it drops. Restarting the agent fixes it at once.

The console’s fleet page shows last seen for an offline robot, which usually says which of these it is.