On April 29, a subset of instances in the Australia (Sydney) region were unreachable from the public internet. The cluster itself remained healthy throughout, and no other regions were impacted.

The root cause was a configuration issue on newly provisioned nodes that prevented pods scheduled on them from receiving inbound external traffic. This also meant the issue surfaced gradually as new nodes came online, rather than all at once, and standard cluster health signals continued to report normal, which delayed detection. Once the root cause was identified, the node configuration was corrected, and affected pods were rescheduled, restoring full access.

To prevent recurrence, the node provisioning process is being updated to validate external reachability before new nodes accept production traffic, ensuring this class of issue is caught automatically going forward. We apologize for the disruption to customers in the Australia region.