Database reliabilityIntermediate
Find database query regressions by release
Which database operations became slower or less reliable after a release?
Compare normalized query families across releases to find slower or newly failing operations.
The dataset and query remain in this browser.
Published expected result
Average duration by query and release
billing_lookup
155 ms
155 ms
billing_lookup
300 ms
300 ms
select_order
133.3 ms
133.3 ms
select_order
633.3 ms
633.3 ms
update_account
87.5 ms
87.5 ms
update_account
140 ms
140 ms
| query_fingerprint | release | operations | average_duration_ms | failed_operations | error_rate_pct |
|---|---|---|---|---|---|
| billing_lookup | 2026.07.1 | 2 | 155 | 0 | 0 |
| billing_lookup | 2026.07.2 | 3 | 300 | 1 | 33.33 |
| select_order | 2026.07.1 | 3 | 133.3 | 0 | 0 |
| select_order | 2026.07.2 | 3 | 633.3 | 1 | 33.33 |
| update_account | 2026.07.1 | 2 | 87.5 | 0 | 0 |
| update_account | 2026.07.2 | 3 | 140 | 1 | 33.33 |
How to read the query
- Query fingerprints group equivalent statements without exposing literal values.
- Release-level aggregates separate application changes from one-off slow queries.
- Latency and error rate belong together: either can identify a harmful regression.
Decisions the SQL cannot make
- 1Investigate fingerprints whose latency or error rate rises in the newest release.
- 2Correlate a regression with deployment and incident timelines before rollback.
- 3Create an alert only after choosing a volume floor that controls low-sample noise.