Some deployments to our GCP (Google Compute Cloud) test and staging environments have been failing now and then with “Node Allocation Failure”. Google Cloud has limited compute capacity in the affected availability zone, so new machines sometimes can’t be provisioned in time. Your application isn’t causing these failures, and production applications that are already running are not affected.
We have made several provisioning improvements that reduce how often this happens. We are also setting up new test and staging zones with more capacity, which we expect to be ready soon. We will update this page when they are.
Until then, retrying a failed run usually works. Using the default resources for test and staging, rather than pinning a specific architecture or machine size, also lowers the chance of hitting the error.
If a change is urgent, you can use “Skip tests” in the Console to deploy to production.
We’re sorry for the disruption. If you are still blocked, contact support@vespa.ai.
