Hi everyone — I’m looking for guidance on a reproducible, project-scoped Compute Engine control-plane failure.
On 10 August 2026, three sequential instances.create requests from the same active, billing-enabled project failed across separate continents:
• us-west4-a at 09:51–09:54 UTC
Error: The service is currently unavailable.
• europe-west4-a at 09:59–10:02 UTC
Error: Internal error. Please try again or contact Google Support.
Correlation code: 658AE6782E509.4FB5C7C.723086C5
• asia-southeast1-b at 10:03–10:06 UTC
Error: The service is currently unavailable.
The US and Asia failures have the same normalized error fingerprint:
cc98c937297aaa5f2bc49d5f2895738f6db936877f56a3cde4dbf5e7ed6204d1
All requests used an ordinary n2-highmem-8 VM with a 100 GB pd-balanced boot disk. Preflight checks passed before each attempt. After every failure, the exact VM name was checked and confirmed absent, so this was not a guest startup, disk-content, or workload failure.
Personalized Service Health currently shows only an apparently unrelated Cloud Storage Insights metadata-snapshot delay, not a Compute Engine incident.
Has anyone encountered multi-regional, project-scoped instances.insert failures like this?