The Google Cloud London Outage (Oct 19-20): What It Pays and How to Claim
October 22, 2026
On the night of October 19, a power event took down europe-west2-b, Google's London zone, for just under eight hours. If your workloads sat in that zone, you are owed credits, and the amounts are already fixed by the SLA even though Google has not yet published its post-incident analysis. This post works out what the outage pays per service and how to file before the clock runs out.
What happened
The incident ran from 23:08 PDT on October 19 to 07:00 PDT on October 20: 7 hours 52 minutes, per Google's status page. The cause was a power event that knocked out the europe-west2-b zone, while other europe-west2 services saw higher error rates and intermittent network issues. More than 35 products are listed as affected, including Compute Engine, GKE, Cloud Run, Cloud SQL, Cloud Storage, Cloud Spanner, Bigtable, Firestore and Pub/Sub. Looker Studio was not affected. Most workloads recovered by 05:10 PDT on the 20th, and Google says a full analysis will follow.
For billing purposes one detail decides who can claim: this was a zonal event. Only europe-west2-b went down, so the credits belong to single-zone resources in that zone and to services whose control planes sat there, not to everyone running in London.
What each deployment shape earns
October has 44,640 minutes. A resource down for the full 7 hours 52 minutes lands at 98.94 percent monthly uptime, and that number crosses different tiers depending on the SLA it sits under.
| Your setup in europe-west2 | SLO | Credit on 98.94% |
|---|---|---|
| Single VM, any family, in europe-west2-b | 99.9% | 10% |
| Single memory-optimized VM | 99.95% | 10% |
| VMs across 2+ zones (all must fail) | 99.99% | none for most setups |
| GKE zonal cluster, control plane in -b | 99.5% | 25% (50% cap) |
| GKE regional cluster (control plane) | 99.95% | 10% if all control planes failed |
| Cloud Run, region-wide >1% 5xx | 99.95% | 10% (50% cap) |
| Cloud SQL Enterprise Plus HA | 99.99% | 10% if all requests failed |
| Cloud Storage regional Standard | 99.9% | 10% if errors were total (50% cap) |
Two rows deserve a closer look. GKE's zonal SLA sits at 99.5 percent, so a 98.94 percent month lands in the 25 percent tier, the highest bracket most people in this outage will see. And Cloud SQL only covers high-availability shapes: single-zone instances are excluded from the SLA entirely, so a lone Cloud SQL instance in europe-west2-b earns nothing no matter how long it was down.
The multi-zone row is where most expectations break. Compute Engine's 99.99 percent commitment only breaches when all of your running instances in the region lose connectivity at the same time. With one zone down, instances spread across zones kept running, so the majority of multi-zone customers have no claim. The claimants are single-zone workloads in europe-west2-b.
The money
Credits are a percentage of the monthly bill for the covered service in the region, applied to future bills. They are not cash. Compute Engine's schedule runs 10/25/100 percent, with the month's bill for that service as the ceiling.
| October spend on the affected service | Credit | Payout |
|---|---|---|
| Single VM, $300/month | 10% | $30 |
| GKE zonal control plane, $200/month | 25% | $50 |
| Cloud Run, $500/month | 10% | $50 (capped at $250) |
| Cloud Storage regional, $100/month | 10% | $10 (capped at $50) |
The percentages sound small until you remember this is per month, per service, per affected region, and October is not the only month this can happen. A 25 percent GKE credit on a control plane you pay for every month is a real line item.
The claim window
Google pays only if you ask, with proof, on time. The windows differ by service: Compute Engine gives you 60 days from the moment you become eligible, while Cloud Storage, Cloud Run, GKE and Cloud SQL all use 30 days. In practical terms that puts most of this outage's claims in November, with Compute Engine stretching into December. Do not wait for Google's post-incident analysis. The SLA terms are already in force, and the clock does not pause while the report is written.
File this month. October's numbers are already fixed, Google's post-incident report will not change the SLA terms, and the 30-day window on most affected services closes in November.
How to file
Use the Cloud Platform SLA contact form (support.google.com/cloud/contact/cloud_platform_sla) and submit once per product per month. Include the project, the region and zone (europe-west2-b), the affected resource IDs, the downtime window in UTC with timestamps, and log files showing the failures. Compute Engine, Cloud Run and GKE all require log files showing the downtime periods and their dates and times. Redact anything sensitive. Google applies approved credits within 60 days of the request.
Do this now
Open the console, find every resource in europe-west2-b, and pull the request logs for the window from 23:08 PDT October 19 to 07:00 PDT October 20. Any of them showing provider-side errors or lost connectivity in that window is a claim. File it before the end of November, and keep the evidence folder for the ones you file in December.