A failed data-quality test is a signal to investigate, not an automatic release decision. Block promotion when a failure breaks an important correctness, integrity, or contractual requirement; allow a lower-risk failure to proceed only when it remains visible, has an owner, and has a tracked follow-up. The commands and severity behavior below are specific to dbt. Other database and orchestration tools need equivalent controls verified in their own documentation.
Decide what should block a release before a test fails
For each check, write down the invariant it protects, which models and downstream consumers depend on it, who owns it, and what the team should do when it fails. dbt’s built-in data tests include uniqueness, non-nullness, accepted values, and relationships. Their presence does not, by itself, determine whether a release must stop.
Reserve a blocking gate for failures that make the published data unsafe or materially misleading—for example, a broken key or relationship assumption on which a critical model depends. A lower-impact anomaly may be advisory if consumers can safely use the release and the issue is explicitly tracked. The appropriate threshold is a policy decision for your system, not a universal cutoff.
- Invariant: What must remain true?
- Impact: Which tables, decisions, consumers, or contractual obligations are affected if it is false?
- Disposition: Must promotion stop, or can it proceed with a visible warning and follow-up?
- Ownership: Who investigates the result and closes the issue?
dbt documents severity and error/warning threshold configuration for tests; see the severity configuration and error threshold references. Those settings implement a team’s chosen policy; they do not define that policy for it.
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →#1 Best Overall
Make warnings visible without making every failure a stop
In dbt, a test can be configured to produce a warning or an error, with thresholds shaping when those outcomes apply. dbt Labs describes warnings as allowing a run to continue and errors as stopping it. The dbt-project-evaluator guide also shows a pattern that warns by default and overrides severity to error in CI using an environment variable: project-evaluator test guidance and dbt Labs’ discussion of data-quality checks.
That pattern is an example, not a universal release rule. Decide whether CI and production should use the same severity based on the consequence of each test failing. A warning must still appear in the pull request or release record; it is not a passing result. Do not silently lower severity simply to make a pipeline green.
Rank #2
Isolate pull-request validation from production
Use pull-request validation to test the change without treating every pre-existing issue in the entire project as a new regression. dbt CI can build and test modified assets and relevant downstream dependencies in a temporary schema, then report status on the pull request. Repository merge protections can require the checks that the team has designated as mandatory. See dbt’s workflow guidance.
- Keep environment targets separate. Maintain distinct development and production targets so validation and experimentation do not write into production.
- Scope PR work to the changed graph. Build and test changed assets and the downstream dependencies relevant to the change in an isolated temporary schema.
- Expose results in the PR. Make warnings and failures inspectable, rather than hiding them behind a pipeline status.
- Require only safety-critical checks for merging. Configure repository protections to gate on the tests the team has classified as mandatory.
- Run broader validation when needed. Selected modified-graph CI is not a substitute for full-project validation where broader assurance is required.
Snowflake’s guidance on using dbt in a pipeline also distinguishes full-project validation from CI focused on the modified graph and recommends integrating checks into the pipeline: Snowflake’s dbt guidance. These are dbt workflow examples; they do not establish portable settings or release behavior for other tools.
Quick wins for a faster PC:
Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Clear out junk files and repair common Windows errorsFree Scan →Investigate the failing rows before choosing a disposition
In dbt, data tests return rows that fail the test’s condition. Inspect both those records and the compiled test query to determine what the check actually evaluated. When the default output lacks useful context, a custom test can return identifying columns. Stored failures can make investigation easier, but a test’s stored results replace that test’s previous results, so copy or retain evidence elsewhere if an incident record must persist. The relevant options are described in dbt’s data-test documentation.
- Confirm which model or source test failed and inspect the compiled query and returned rows.
- Check whether the failure is reproducible and whether it is newly introduced by the change.
- Determine whether changed transformation logic, source data, or an execution/configuration problem explains the result.
- If an upstream source load is stale or changed, refresh or correct that input and rerun the relevant check.
- Choose a disposition based on the affected invariant and impact, then record the owner and next action.
A failure may be unrelated to modified or errored nodes. dbt’s workflow guidance describes, for example, a source test failure that may require a refreshed load. Diagnose and document that case rather than silently waiving it. The same guidance describes selecting failed tests and excluding a known example; use such exclusions only with a named reason and review, not as a blanket way to suppress inconvenient results.
Rank #4
- HP ProLiant DL360 G7 8B Server
- 2x X5650 2.66GHz 12-Cores Total
- 32GB RAM / 8x 146GB 10K 2.5in SAS Hard Drives
- P410 w/ 512MB
Record exceptions so advisory failures do not become permanent
The following disposition record is an operational recommendation, not a vendor-prescribed schema. It keeps a low-risk release decision reviewable and gives remediation a clear owner.
- Test name and affected model or table
- Failing-row count or representative sample, if available
- Likely cause and whether the issue is related to the change
- Severity and release decision
- Rationale and approver for any exception
- Owner and due date for remediation
If a low-risk failure proceeds, retain its warning in the PR or release record and track the follow-up. If a high-impact invariant fails, stop promotion unless an authorized, documented exception is permitted by the team’s policy. Avoid broadly disabling tests or excluding failures without a specific reason, owner, and review.
Apply the same risk logic outside dbt, but verify the controls
The release principles—classify impact, keep results visible, isolate validation, inspect evidence, and assign exception ownership—apply broadly. The specific severity settings, temporary-schema CI behavior, failure-row handling, and exclusion mechanisms described here are dbt-specific. The cited material does not establish equivalent syntax, rollback behavior, or release gates for every database, orchestrator, or testing framework. For another tool, verify its documentation before assuming that a warning continues execution or that a scoped check isolates changes in the same way.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
