ReliabilityAdvanced
Measure job recovery by customer plan
Which failed jobs recovered on retry, and which accounts still need intervention?
Collapse attempt-level worker events into one logical job before comparing recovery and unresolved failures by plan.
The dataset and query remain in this browser.
Published expected result
Unresolved logical jobs by plan
starter
2
2
growth
0
0
business
0
0
free
0
0
| plan | logical_jobs | jobs_with_failure | recovered_jobs | unresolved_jobs |
|---|---|---|---|---|
| starter | 2 | 2 | 0 | 2 |
| growth | 2 | 1 | 1 | 0 |
| business | 1 | 0 | 0 | 0 |
| free | 1 | 0 | 0 | 0 |
How to read the query
- Attempts are grouped by logical job identifier before failure and recovery are counted.
- A job may have both a failed attempt and a completed retry, which is recovery rather than permanent failure.
- Joining after the reduction prevents a retried job from multiplying account-level counts.
Decisions the SQL cannot make
- 1Define terminal job statuses and the maximum expected retry delay.
- 2Distinguish automatic retry recovery from manual replay.
- 3Alert on unresolved work and age, not every transient failed attempt.