Gateway Retention, Quota and Limits

Updated Jul 26, 2026
Hosted File Gateway

Gateway Retention, Quota and Limits

The line under the counters on the Transfers tab answers both of the questions an operator actually has while looking at the list: when does this go away, and is there room for the next one.

The retention and storage line: files kept for 7 days, the option that sets it, and storage used

Retention

Every file a partner sends stays in your bucket as the evidence of what actually arrived — which is the point of a receipt, and also means the bucket grows forever unless something removes them.

Retention uses the platform's ordinary TTL options, with one addition: a TTL option carries a flag saying it is the one that governs gateway files.

The TTL option dialog with the Applies to gateway files checkbox highlighted
Company TTL Option → Applies to gateway files. Flag one option; if none is flagged, a built-in default applies.

How the window is chosen

  1. Your company TTL options are searched for an enabled one with the gateway flag.
  2. Otherwise the platform options are searched for the same flag. DataMagik ships one called Gateway Received Files, defaulting to 7 days.
  3. Otherwise a built-in default of 7 days applies.
The platform Gateway Received Files TTL option card with its Gateway files badge
It is matched on the flag, not on the name. Renaming the option is safe. Creating a company option with a similar name but no flag is not — it would be ignored while the screen still showed a plausible window.
Gateway files do not inherit your default document TTL, on purpose. A document-retention policy — often a day or two for generated documents nobody keeps — is a terrible policy for the evidence of what a trading partner sent. It is a separate decision.

Changing it

Press Change it on the retention line, or go to the TTL dashboard. Platform options are read-only to you; to use a different window, create a company option with the gateway flag ticked. Yours wins over the platform's.

An option with auto-cleanup switched off means keep indefinitely — an explicit choice to retain is not something a sweep should quietly override. Minimum and maximum days clamp the window.

What cleanup actually removes

The scheduled sweep deletes the stored object and marks the receipt purged. The row stays: an operator asking what became of a file deserves "deleted on schedule" rather than silence. The expiry date is shown per transfer in the list.

Storage quota

Gateway files count toward your company storage allowance alongside documents and lookup data. The retention line shows what is used and against what limit.

SituationWhat happens
Your plan has no storage limitUnlimited.
Paid plan, over the allowanceAccepted and recorded as overage. The allowance is a billing threshold, and refusing a paying customer's partner at the door is a support incident, not enforcement.
Paid plan, more than 10× the allowanceRefused. A real business that outgrows its tier lands between one and two times it. Ten times means something is broken — most likely a partner's client stuck in a loop — and continuing would keep billing you for a fault nobody has noticed.
Free plan, over the allowanceRefused. There is no billing relationship for the overage to land in.

A refusal is a clean protocol error to the partner and a refused transfer recorded for you, never a silent drop. Warnings appear at 80% and 95%.

The check fails open on any internal error. Losing an EDI delivery because a usage table was briefly unreadable is a far worse outcome than storing a few megabytes that should have been refused.

Duplicate detection

Every upload is hashed. If the same principal delivers the same bytes again, the transfer is recorded as duplicate, the note names the original, and no handler runs.

  • The comparison is content only — a re-upload under a different filename is still a duplicate.
  • It is scoped per principal. Two partners sending identical files are not duplicates of each other.
  • The bytes are kept either way, because they are the evidence of what was sent.
If a partner legitimately re-sends identical content that you do want processed again — the same daily file with no changes — use replay on the original transfer rather than asking them to resend.

Transfer and connection limits

LimitValue
Maximum size of a single transfer256 MB
Concurrent SFTP sessions200
Concurrent FTPS sessions100
Idle timeout15 minutes
Concurrent passive FTPS data transfers20 (the declared passive port range)

Uploads are streamed to storage rather than held in memory, so a large file does not depend on the service having room for it.

Connection throttling

A source address that accumulates 25 failed authentication attempts within 15 minutes is throttled. The check runs before any credential work, so a brute-force run costs a counter lookup rather than a password comparison.

What the partner experiences

Connections are refused at authentication. To their client it is indistinguishable from a wrong password — there is no message explaining they are throttled, deliberately, since that message would also be a hint to an attacker. It clears on its own as the failures age out of the 15-minute window.

The usual cause is not an attack. It is a partner whose scheduled job runs every minute with a stale password or a rotated key. Check Recent connection attempts on the Security tab: a repeating username from one address is a misconfigured job, while many different usernames from one address is someone guessing.

To get a partner working again, fix the credential — then have them stop and restart their job rather than letting it keep failing, so the window can clear.

Throttling fails open if the audit log cannot be read: a database blip must not lock out every partner at once. The per-credential checks still apply.
Was this page helpful?