DBLift emits four kinds of HTML report. They share a header, a verdict, a KPI band and a findings table, so once you can read one you can read all of them. Every report prints to white, so an approver can file it as evidence.
The four reports
| Report | Answers | Edition |
|---|---|---|
| migration | What actually ran, statement by statement, and how long it took | OSS |
| validate-sql | Which policy rules the SQL broke, and where | Pro |
| plan | What is pending against a snapshot, and whether anything drifted | Enterprise |
| preflight | Whether the release is ready, including a rehearsed replay | Enterprise |
Anatomy of a report
This is a real preflight report, annotated. Read it top to bottom: verdict, then what drove it.
Header and verdict. The timestamp is the report's identity — the same one appears in every artifact filename from that run. Fail on error tells you which threshold produced this verdict, so a PASS is only as strong as the threshold beside it. The snapshot line names the exact artifact the run was judged against.
KPI band. Counts, not judgements. Zero pending and zero drift with a FAIL verdict means the failure came from somewhere else — phase failures or SQL findings. Replayed scripts is the one to check on preflight: a zero there means nothing was actually rehearsed.
Findings. Every row carries a stable rule id, a file-and-line location, and a message. The rule id is what you cite in a review or suppress in a rules file — the message wording may change, the id does not.
Severity and verdict
Severity describes a finding. The verdict describes the run, and it is decided by --fail-on. A report full of warnings still says PASS when the threshold is error.
| Severity | Meaning |
|---|---|
| error | Blocks the release. Checksum drift, snapshot integrity failures, SQL policy errors. |
| warning | Needs a human decision. Large backfills, transaction-boundary assumptions. |
| info | Context for the record. Pending migrations, snapshot provenance. |
Keeping reports as evidence
Request several formats in one run and archive the directory from CI. The HTML is for people, the JSON for machines, and every file shares the run's timestamp so the set stays together.
dblift preflight --env uat --skip-replay --format json,html --output-dir dblift-reports --fail-on error[!NOTE] Reports print
The HTML report inverts to white with light table headers when printed or saved as PDF. Nothing in a report depends on hover or colour alone, so a printed copy carries the same information as the screen.