Run a pipeline
Run one imported tap-target pair at a time. PipelineWise selects FastSync for an eligible initial load and the Singer path for ongoing incremental or log-based replication.
Preflight
Check configuration, connectivity, and imported status before starting a load:
pipelinewise validate --dir ./pipelinewise_samples
pipelinewise test_tap_connection --tap orders --target snowflake
pipelinewise status
Re-run import_config after changing project YAML. Validation alone does not
update generated runtime configuration.
Start and observe
pipelinewise run_tap --tap orders --target snowflake
During the run, PipelineWise:
locks the tap-target pair to prevent a concurrent run;
selects FastSync-eligible initial tables and Singer tables;
streams or stages records into the target;
persists only target-acknowledged state; and
marks the log
successorfailed.
Follow the active log from another terminal:
tail -f ~/.pipelinewise/snowflake/orders/log/*.running
See Logging for filenames and diagnostic collection.
Stop safely
Use the CLI rather than killing an arbitrary child process:
pipelinewise stop_tap --tap orders --target snowflake
The tap stops producing records and the target is allowed to finish data it has already received. For PostgreSQL LOG_BASED replication, WAL feedback remains bounded by target-acknowledged state. An unexpected termination can replay records after restart; targets must therefore retain their normal primary-key merge semantics.
Confirm success
After the command exits successfully:
pipelinewise status
Confirm the target data as well as the PipelineWise status. For critical tables, configure Data-diff checks rather than treating a successful process exit as proof of source-to-target equality.
Recover from failure
Do not advance or edit state merely to make a failed run start. Instead:
retain the failed log and exact error;
correct the source, target, credential, or configuration problem;
restart the same
run_tapcommand; andverify target contents and the new acknowledged state.
Use Troubleshooting for known errors. Use Resync and repair only when the required source log or replication slot is no longer available, or when target data must be rebuilt deliberately.
Schedule the command only after an interactive run succeeds; see Scheduling.