Status

Single node, single region. This page is written by the deploy job, so it is as fresh as the last release — it is not a live health probe. For machine-readable health use /v1/healthz from inside the network.

Maintenance

Restarts for updates happen on Sundays around 03:00 UTC and take under a minute. In-flight warms resume on the next request; API calls during the window get a 503 with Retry-After.

Known issues

Objects above the per-object ceiling silently skip caching and pass through to origin instead of erroring — a quieter failure mode than most tools.

Purge during an active warm for the same key can occasionally leave one stale byte-range response in flight; it expires on its own within the TTL.

Contact

Operational problems: the address is on the privacy page.