Re-running only what failed with dbt retry

The dbt retry command re-executes only the nodes that failed or were skipped in the most recent dbt invocation, using the run_results.json file to determine what to rerun. This saves time in development and CI by avoiding a full re-run of successful models, tests, or seeds. It works with any dbt command that produces run_results.json, such as run, build, test, or seed.

1 · Learn the must-know

• 1dbt retry reads the run_results.json artifact from the prior invocation to identify nodes with a status of error, fail, or skipped.
• 2It reruns the failed/skipped nodes using the exact same command and flags (e.g., full-refresh, target, vars) as the original invocation that produced run_results.json.
• 3If no run_results.json is found, or if the prior run had zero failures, dbt retry will raise an error since there is nothing to retry.
• 4dbt retry does not accept its own selection criteria (like --select or --exclude); it strictly reuses the failed/skipped node set from the previous run.
• 5You must run dbt retry from the same project directory (and target/profile context) as the original failed invocation for the artifacts to be valid.
• 6Successful nodes from the prior run are left untouched and are not rerun, which distinguishes retry from simply rerunning the full original command.

2 · Check your understanding