Skip to content

Cloud · Google

Google Cloud europe-west2 (London) power loss: utility cable fault during single-feed maintenance, generators never started

Started
Oct 20, 2026, 06:00 UTC
Resolved
Oct 20, 2026, 14:03 UTC
Duration
8h 3m (from provider timestamps)
Affected
Compute Engine, Persistent Disk / Hyperdisk, GKE, Cloud SQL, AlloyDB, Bigtable, Cloud Batch, Cloud Build, Cloud Storage, Cloud Workstations, Composer, Dataproc, Filestore, Managed Lustre, Migrate to VMs, NetApp Volumes, Pub/Sub Lite, Vertex AI Workbench, VMware Engine, BigQuery, Pub/Sub, Dataflow, Memorystore, Apigee, VPC
Cause
Final reviewThe London facility hosting europe-west2-b was running on a single utility feed because feeder B was open for repairs after a generator load test. An underground cable fault on the utility network then cut feed A, and the generator controller never started the generators because its sensors read live voltage upstream of the open feeder B breaker. UPS batteries ran out, the zone lost power, and failover load and cross-zone control-plane calls spread errors across the region.
Was it an attack?
Not stated
State
resolved

Anatomy

  1. 1. Trigger

    Power failure

  2. 2. Already broken

    Missing validation, Generator or UPS failure, Single point of failure

  3. 3. What made it spread

    Control-plane dependency, Shared region, Recovery backlog

  4. 4. Blast radius

    single region: europe-west2 (zone europe-west2-b; regional knock-on in -a and -c)

  5. 5. Cause category

    Power or cooling

Labels follow the root-cause taxonomy. A person's change is recorded as the trigger and the missing safeguard as what was already broken; “human error” is never a cause here.

How the explanation changed

  1. October 20, 2026 · Preliminary cause

    “A power event caused an outage within europe-west2-b zone.”

    Google Cloud Service Health incident Xs2XTHzAVC9YnafnPyzj (preliminary 22 Oct, incident report 27 Oct, revised 4 Nov)

  2. October 22, 2026 · Preliminary cause

    “The Programmable Logic Controller (PLC) automation system (both primary and backup PLCs) failed to initiate the logical automatic start/transfer operations of the generators, preventing backup generators from assuming the load.”

    Google Cloud Service Health incident Xs2XTHzAVC9YnafnPyzj (preliminary 22 Oct, incident report 27 Oct, revised 4 Nov)

  3. October 27, 2026 · Final review

    “Following the loss of primary utility feed A, a voltage-sensing design flaw and PLC misconfiguration prevented the facility’s generator plant from receiving an automatic start signal to assume the critical load.”

    Google Cloud Service Health incident Xs2XTHzAVC9YnafnPyzj (preliminary 22 Oct, incident report 27 Oct, revised 4 Nov)

What the provider committed to change

  • Utility provider to locate (done 26 Oct) and repair the cable fault on feed A (est. 13 Nov)
  • Return the facility to primary feeds A and B (est. 17 Nov)
  • Generator load transfer tests through PLC signals (done 20 and 26 Oct) and again after feed restoration (est. 17 Nov)
  • PLC reconfigured and validated (done 26 Oct); full PLC code review (est. 6 Nov); validate PLC configurations at similar sites (est. 15 Nov)
  • Updated emergency operating procedures, practised twice per shift (done)
  • PLC code reviews during commissioning and integrated system tests (est. 18 Dec)
  • Freeze planned maintenance at similar sites and add approvals for out-of-standard changes (done 30 Oct)
  • Automated alert-to-communication for location-impacting incidents; expanded platform testing for infrastructure events

Post-incident reviews

Sources

  1. Google Cloud Service Health incident Xs2XTHzAVC9YnafnPyzj (preliminary 22 Oct, incident report 27 Oct, revised 4 Nov) (provider, read November 17, 2026)
  2. Google Cloud incidents.json feed (same incident, machine-readable timestamps) (provider, read November 17, 2026)
  3. The Register, 2026-11-05 (news, read November 17, 2026)

Last verified November 17, 2026. Spot an error? The methodology page says how corrections work.