On Saturday 3 October 2026, between 10:49 and 17:40 UTC, some document processing jobs failed and document file downloads were unavailable. Affected jobs used certain country compliance integrations, document conversions and delivery steps. The issue is fully resolved, and jobs have processed normally since 17:40 UTC.
The Invopop API and the Console stayed available throughout.
Jobs that reached an affected step during the window failed, and were marked as errors rather than completing.
Affected: steps that read or save documents from our document processing services. These included several country compliance integrations (for example Spain, France, Italy, Poland, Portugal, Argentina, Saudi Arabia and Mexico), Peppol, document import and conversion, email delivery, and some accounting and billing integrations.
File downloads unavailable: download links for document files on silo.invopop.com didn't work, because that endpoint was still served from the older environment. It now points to our new infrastructure.
Not affected: the Invopop API, the Console, and integrations that run on our newer infrastructure.
If a job failed between 10:49 and 17:40 UTC on 3 October, please contact support before retrying it, especially if it submits to a tax authority. We'll help you identify and safely retry affected jobs.
Time | Event |
|---|---|
10:49 | Document processing jobs on some integrations start to fail, and file downloads become unavailable |
17:29 | Cause identified |
17:34 | Fix deployed |
17:40 | Service fully restored; jobs processing normally |
A configuration gap during an infrastructure migration caused one of our document services to stop in part of our platform.
We're moving our services to new infrastructure. On 1 October we released an update to our document service that needed a new database connection. The connection was configured on the new infrastructure, but not on the older environment that still serves some integrations. For two days a previous version of the service kept those integrations working. On 3 October at 10:49 UTC that previous version stopped serving, and the integrations could no longer reach the document service until we added the missing configuration.
We're making three changes so that a release like this can't fail quietly again:
Faster detection. We're adding new alerts that page our on-call engineers as soon as a service fails to start or loses capacity, instead of waiting for errors to build up.
Finishing the migration. We're completing the move of the remaining integrations to our new infrastructure, so every service runs from a single, consistently configured environment.
We're sorry for the disruption this caused.
