Recover SSH or Orca connectivity
Scope and symptoms
Section titled “Scope and symptoms”Mac cannot reach Shipyard, or Orca reports a disconnected worker. No service restart is implied.
Diagnose
Section titled “Diagnose”On the Mac, try ssh -o BatchMode=yes -o ConnectTimeout=8 shipyard hostname. If this fails, inspect the existing SSH alias and Tailscale client status locally. Do not publish complete SSH configuration or key paths from unrelated hosts. Compare normal SSH with Orca’s connection state.
If normal SSH works but Orca’s CLI reports EPERM in a sandbox, retry with the environment’s supported permission mechanism. EPERM is not proof of an Orca outage. Use the existing Orca host entry; do not create duplicate hosts or replace SSH configuration.
Repair and verify
Section titled “Repair and verify”Reconnect the existing Orca host. Check the selected execution host and repository path. Then read task status on Shipyard. Reload the noVNC tab if only the viewer failed. A recorded disconnect/reconnect preserved services. Controlled AC return passed for SSH, Tailscale and Docker; Orca reconnection, task readiness and off-LAN access need separate checks.
Escalation
Section titled “Escalation”If normal SSH cannot connect and Tailscale/network state cannot be resolved read-only, record the error and wait for human network/console access. Do not reboot, change firewall/SSH authentication or restart workloads as a guess.