Verifying what runs unattended.
WHY IT IS NEGLECTED
It has no interface, and failures are invisible.
WHAT TO TEST
That the job runs when scheduled That it processes what it should That it handles failure of individual items That it can be run twice safely That it completes within its window
WHY RUNNING TWICE SAFELY MATTERS
Retries and overlapping runs occur.
WHAT TO TEST ABOUT PARTIAL FAILURE
That one failing item does not stop the rest, unless intended.
WHAT TO VERIFY
That failures are recorded, and surfaced.
WHAT TO TEST ABOUT TIMING
Behaviour when a run takes longer than its interval.
WHAT THAT CAUSES
Overlapping runs, which may corrupt data.
WHAT PREVENTS IT
A lock, verified by testing.
WHAT TO TEST ABOUT VOLUME
Behaviour with far more items than usual.
WHY
Backlogs occur after outages.
WHAT TO TEST ABOUT EMPTY INPUT
That it completes without error.
WHY
Empty cases are frequently unhandled.
WHAT TO VERIFY IN PRODUCTION
That the job actually ran, and produced plausible output.
WHAT TO ALERT ON
A job not running, which is the commonest silent failure.
