n8n Data table as a Sume job ledger: the 200 MiB default limit

Create an n8n Data table with one row per Sume job, insert a row when you submit, and use the Data Table node's Upsert operation to update it when the webhook or a poll reports a terminal status. n8n's docs say all data tables in an instance share a default limit of 200 MiB, and that inserts and updates fail once it is full, so keep rows small: ids, status and result URLs, never video bytes.

The limits are from n8n's Data tables page and the Data Table node page (both read 2026-10-02); the job fields from Sume's Jobs and results and Webhooks pages.

What does the Data Table node offer?

Its Row resource can insert, get, update, delete and upsert rows, and it has If Row Exists and If Row Does Not Exist checks. The Tables resource creates, lists, updates and deletes tables. n8n also says a Code node cannot read data table values directly, so any lookup has to go through the node itself.

What columns does a job ledger need?

Keep one key you control and one key Sume controls. The submit response carries the job id as request_id and the status_url and result_url; the terminal webhook carries event, job_id, a status of OK or ERROR, and payload.artifacts[].url on success.

What happens when the table fills up?

n8n says a warning shows at 80% of the limit and a final warning at the limit, after which manual additions are disabled and workflow executions error when they try to insert or update. Self-hosted instances can change the cap with the N8N_DATA_TABLES_MAX_SIZE_BYTES environment variable.

For a Sume ledger the risk is in the receiver. If the workflow that handles the webhook fails on the upsert, Sume's docs say network errors and non-2xx responses are retried, up to 10 attempts at a fixed 30-second spacing by default, with a 10-second timeout per attempt. Whether your failure reaches Sume as a non-2xx depends on when your Webhook node responds. Either way the job itself still finishes, and POST /v1/jobs/{job_id}/webhook/redeliver re-sends the terminal event once you have freed space.

When should I use something else?

n8n calls data tables suitable for light to moderate storage. If you submit thousands of jobs a day and keep them for months, move finished rows to a sheet or database and keep only open jobs in the table. If you only need to find one execution later, the Execution Data node is lighter.