Status

All systems operational

DeliveryOperational Cached and origin-fetched responses from the edge.
90 days ago99.98%today
Image transformsOperational Resize, crop, re-encode on first request.
90 days ago99.97%today
Video transformsDegraded Progressive renditions and HLS ladders. Queue is longer than usual.
90 days ago99.71%today
UploadsOperational Direct and resumable sessions.
90 days ago100%today
Management APIOperational Purge, tokens, usage.
90 days ago99.94%today

Recent incidents

7 Aug 2026, 18:40 UTC

Video transform queue backed up

A batch of long masters uploaded within a few minutes filled the encode queue, and HLS ladders for new sources started taking up to nine minutes instead of the usual forty seconds. Delivery of already encoded renditions was unaffected. We are adding encode capacity and will publish a follow-up once the queue has drained.

Monitoring · started 18:40 UTC

21 Jun 2026, 09:12 UTC

Purge acknowledged but not applied

For eleven minutes, prefix purges returned 202 without reaching part of the edge, so stale derivatives kept being served for affected prefixes. Single path purges were fine throughout. The cause was a queue consumer that stopped after a deploy without failing its health check. Purges are now acknowledged only once the fan-out is accepted, and the health check tests the consumer rather than the process.

Resolved · 11 minutes

3 May 2026, 22:05 UTC

Elevated errors on AVIF output

An encoder update rejected images with an unusual colour profile, returning 422 for roughly 0.4% of first requests. Cached variants and every other output format were unaffected. We rolled the encoder back within the hour and now hold a corpus of awkward source files that runs before an encoder ships.

Resolved · 52 minutes