run
Validate and execute a pipeline — once, on CRON, or as a long-running listener.
cannectors run [--dry-run] <pipeline.yaml>Reads the pipeline, validates it (same checks as
validate), resolves environment variables,
then executes it. The process lifetime depends on the input.
Examples
# Run once and exit
cannectors run ./examples/01-http-polling-basic-to-http-batch.yaml
# Preview without writing to the output
cannectors run --dry-run ./examples/19-http-output-single-template.yaml
# Stay alive on CRON
cannectors run ./examples/04-http-polling-cursor-oauth2.yamlLifetime
| Input | Lifetime |
|---|---|
httpPolling without schedule | One-shot. Exits 0 on success. |
httpPolling with schedule | Stays alive. Triggers on every CRON match. Exit on SIGINT/SIGTERM. |
database without schedule | One-shot. |
database with schedule | Stays alive on CRON. |
webhook | Long-running HTTP listener. Exit on SIGINT/SIGTERM. |
The runtime drains in-flight work and persists state on shutdown — a graceful stop never loses a batch.
Dry-run mode
--dry-run runs everything except the output's side effect: the
input fetches, every filter executes, then the would-be output
payload is printed to stdout instead of being sent.
cannectors run --dry-run sync-orders.yamlSee Dry-run mode for the full mental model
and dryRunOptions tuning.
Flags
| Flag | Effect |
|---|---|
--dry-run | Skip the output side effect; print a preview instead. |
--verbose | Per-record trace lines on stdout. |
--quiet | Suppress informational logs (errors still print). |
--log-file <path> | Write structured logs to a file. |
See Flags for the full reference.
Exit codes
| Code | Meaning |
|---|---|
0 | Pipeline completed successfully (one-shot mode) or was stopped cleanly. |
1 | Validation errors before execution. |
2 | Parse error. |
3 | Runtime error — a stage failed past its onError policy. |
Production process management
The run process is single-replica by design. Don't run two copies
of the same scheduled pipeline — they'd fire on the same CRON tick
and double-process records. Use a single-replica Deployment,
StatefulSet, or systemd unit.
See Operations · scheduling in production for the deployment patterns.