Start with the agent’s own checks
On the node, run the diagnostics for the area that looks wrong:ados diag video and ados diag link sample cumulative counters twice and report which hop stopped moving. A single snapshot of a counter cannot tell a stalled flow from a healthy one, so trust these verdicts over a reading you took yourself.
Then read the journal. It is the log of record, because the Black Box store is off by default (logging.store.enabled: false) and ados logs returns nothing until you turn it on:
journalctl -u 'ados-*' --no-pager | grep factory_reset_performed. Where you have enabled the store, ados logs query and ados logs tail search the same events with telemetry alongside; see Black Box.
If the node does not answer at all, go to Recovery first.
By area
Agent diagnostics
Per-hop verdicts for video, the radio link, and storage wear.
Drone Agent troubleshooting
Flight controller not detected, video, radio link, cloud, and an agent that will not start.
Ground Agent troubleshooting
No video, access point, reaching the station, USB tether, and the display.
Mesh troubleshooting
Role transitions, mesh pairing, gateway choice, and partitions.
Vision Navigation troubleshooting
Optical-flow quality, EKF source switching, VIO drift, and scale.
Developer troubleshooting
Plugin exit codes, signatures, manifests, capabilities, and IPC.
Self-hosting: troubleshooting and ports
A self-hosted cloud stack and the ports it needs.
Recovery
Reach, repair, reset, or reinstall a node that stopped answering.
Still stuck
Collect a support bundle on the affected node before rebooting it, then open an issue or ask in the community. Support escalation covers what the bundle holds, what to redact, and where to send it.Next steps
Recovery
Reach and rebuild a node.
Support escalation
Collect a bundle and report a fault.