Closed general

Silkscreen dashboard stop returns before its server lease is released

John Lauer · ·closed by John Lauer

During maintenance verification, sequential commands for the same task-owned run raced:

python3 tools/silkscreen-dashboard/server.py stop --run /home/adom/aiflow-maintenance-20260917 --ai-thread "ESC AI Flow Astra" --reason "Reload task-owned test server with encoded-frame validation"
python3 tools/silkscreen-dashboard/server.py start --run /home/adom/aiflow-maintenance-20260917 --ai-thread "ESC AI Flow Astra" --port 45813

Full command output:

{"ok": true}
ERROR: server did not become healthy; see /home/adom/aiflow-maintenance-20260917/silkscreen-dashboard/server.log

The server log says OK: another instance for this run/thread is starting or running. Stop removed the registry before the old process released its exclusive lease. Start spawned its replacement during that interval, so the child exited. A later start succeeded; no shared process was killed.

Expected: stop returns after its task-owned lease is released, or explicitly reports shutdown still pending. Fix in progress: wait up to ten seconds for that exact lease, then allow a sequential restart. No broad PID operations. This is a lifecycle failure, not a solver failure. Board/spec: /home/adom/aiflow-maintenance-20260917/board-0.kicad_pcb and /home/adom/aiflow-esc-astra/spec.json.

1 Reply

John Lauer ·

Published and installed AI Flow 0.1.28. Stop waits for its exact task-owned lease to release. Three consecutive real stop/start cycles completed in 0.86, 0.84 and 0.82 seconds. No broad process operation. Source, installed registry payload and active binary hashes agree.

John Lauer closed this ·

Log in to reply.