Retrying & re-running

Runs are final — once a run reaches Succeeded, Failed, or Canceled, it never transitions again. To recover from a failure or push more data, start a new run.

What carries forward between runs

Three things persist, which is why re-running is cheap:

• The extracted data from the source.
• The staged records ready to load, with their pending-change flags and their Shopify identity once loaded.
• The mapping version pinned to each resource.

A new run picks up where the data is, not where the previous run stopped. There’s no “resume” button — Graftport just starts processing the phases you ask for, against the current state.

Recovering from a failed load

A load run failed mid-way. Some records loaded successfully, some failed, some never started.

To finish:

1. Check the failed run item’s errors. If they’re rate-limit or transient, just start a new load run. Only-dirty is on by default, so the new run only touches what didn’t load.
2. If the errors are payload problems (e.g. invalid SKU format), fix the mapping, run a transform-only pass to restage the affected records, then load.

Recovering from a failed transform

A transform error usually means the mapping crashed on a specific row. The run item logs which row broke and the JSONata error. Fix the mapping in the editor, save, and run transform again — the rows that already transformed successfully will retransform (cheap), and the broken rows will succeed this time.

Recovering from a failed extract

Extract failures are almost always credentials, rate limits, or network. Look at the error on the run item:

• 401/403 — bad credentials. Update them in the migration’s settings, then start a new extract.
• 429 — rate-limited; Graftport backs off automatically, but if the source is throttling Graftport specifically, contact the source platform’s support.
• 5xx — source platform issue. Wait, then start a new extract.

Manual retry

The Retry button on a failed run item starts a new run with the same configuration as the original, scoped only to the failed resource type. Faster than reconfiguring from scratch when you know the original setup was right.

Canceling a run

Hit Cancel on a running run to stop it after the current batch. Cancel is graceful — Graftport finishes the in-flight call and then marks the run Canceled. Records loaded up to that point keep their Shopify identity and won’t be re-pushed unnecessarily.