Google Cloud's London zone went down for almost eight hours
A power event in europe-west2-b took out dozens of Cloud products overnight on 19 to 20 October, and some of the knock-on outages at platforms built on top of it lasted longer than the incident itself.
Google Cloud’s europe-west2 region, which runs out of London, went down overnight on 19 to 20 October. The incident opened at 23:08 Pacific time on the 19th and was marked fixed at 07:00 Pacific on the 20th, a run of almost eight hours. Google’s own status page gives the cause as “a power event” within the europe-west2-b zone, since mitigated.
The affected list runs to more than 35 products, not just compute: Compute Engine, GKE, Cloud SQL, Spanner, AlloyDB, BigQuery, Cloud Storage, Pub/Sub and several Memorystore tiers were all named. Platforms that run on that zone felt it too. MongoDB Atlas, JFrog’s cloud platform and the hosting provider Hostinger each told their own customers the disruption traced back to this Google Cloud incident.
# if you run production in europe-west2, worth checking now:
gcloud compute instances list --zones=europe-west2-bWhy it matters: a single zone losing power took down managed databases, load balancing and networking at the same time, which is exactly the blast radius multi-zone deployment is meant to contain. If your infrastructure, or a vendor you depend on, sits only in europe-west2-b, this is the incident to point at when you ask for budget to spread across zones.
The caveat: Google’s report names the europe-west2-b zone specifically, not the whole europe-west2 region, so how much this affected any one customer depended on which zone their resources were actually in.