FIELD NOTE / TIKTOK
The backfill job ran. The reservation was late.
The complete written thought and the evidence behind it. The video edition will follow its public release.
The written argument is here.
This approved TikTok edition is on the journal now. Its video player and original platform link will appear after each public release is verified.
The backfill job ran. The reservation was late.
Video caption
The backfill job ran. The reservation was late. The broken promise may predate the small job. Scheduler rule and customer promise both need evidence. #EricFieldNotes
Full written post / accessibility read
Watch this timeline. The customer was promised a noon start. Slurm estimated one o'clock from current jobs and time limits. A short backfill job finished before one. The large run began after noon. Blaming the small job would be a bad diagnosis.
Slurm's documented backfill rule checks that a lower-priority job should not delay the expected start of higher-priority work. In this scenario the estimate was already later than the customer's contract. The root cause could be the reservation model, a time limit or topology.
Recreate the queue in a disposable scheduler. Run it with and without the backfill job, keeping the original time limits and topology. Compare predicted and actual starts. Then vary the suspect estimate. The result identifies whether backfill, estimation or contract setup explains the miss.
Do not disable useful backfill because a late reservation happened nearby. Repair the cause shown by the counterfactual, and separately gate the external start-time promise. Do this because an optimized queue is valuable only when the promised service is still delivered.
#EricFieldNotes
Four-beat scene transcript
1. The backfill job ran. The reservation was late.
Watch this timeline. The customer was promised a noon start. Slurm estimated one o'clock from current jobs and time limits. A short backfill job finished before one. The large run began after noon. Blaming the small job would be a bad diagnosis.
Visual: Those two facts do not prove one caused the other.
2. Proper backfill preserves expected start.
Slurm's documented backfill rule checks that a lower-priority job should not delay the expected start of higher-priority work. In this scenario the estimate was already later than the customer's contract. The root cause could be the reservation model, a time limit or topology.
Visual: The broken promise may predate the small job.
3. Remove the small job in a replay.
Recreate the queue in a disposable scheduler. Run it with and without the backfill job, keeping the original time limits and topology. Compare predicted and actual starts. Then vary the suspect estimate. The result identifies whether backfill, estimation or contract setup explains the miss.
Visual: Then inspect the scheduler's reservation trace.
4. Keep two success tests.
Do not disable useful backfill because a late reservation happened nearby. Repair the cause shown by the counterfactual, and separately gate the external start-time promise. Do this because an optimized queue is valuable only when the promised service is still delivered.
Visual: Scheduler rule and customer promise both need evidence.
Research and claim limits
The examples identified as illustrative or simulated are design probes, not reported incidents. Vendor specifications do not establish workload performance.