Situation: A compute job failed or stalled

Triggers: A *-compute background job (build, mine, refine, raid) exited, produced no completion, or has been running far longer than expected.

60-second diagnosis

cat memory/jobs/<job>.log      # what did the compute helper print?
ps -p "$(cat memory/jobs/<job>.pid)" 2>/dev/null && echo running || echo not running

• Process not running, no completion tx → it errored out. Read the log tail.
• Completion broadcast but state unchanged → the completion failed a precondition (charge, power, shield window). Broadcast ≠ success.
• Still running after many hours → normal at low -D (the wait is the age clock), unless the log shows it started hashing and is grinding.

Common causes and fixes

Stop / escalate

• Never launch a second compute on the same signing key while one is in flight.
• If a long build/raid keeps failing its window, stop burning proof-of-work and reassess the target/timing.

See also

• async operations · hashing
• Skills: production · building · combat