At 06:08 UTC on Tuesday 20 October, a power event knocked out part of Google Cloud's europe-west2-b zone in London. Google's incident lists 38 products, from Compute Engine to Cloud SQL and BigQuery, and ends at 14:00.
Google's status page said nothing about it until 08:35.
We follow about 9,800 official status pages. That day, 70 of them reported a problem in words that point back to London. Here's the day as they told it:
20 Oct, UTC 08:35
Status pages that said it 50 /70
Still reporting 50
Google Cloud's own page Major outage since 08:35
08:35 Google Cloud posts its first update. 50 pages already had.
Power event Google posts Mitigated Google closes
06:00 12:00 Midnight
Named Google Cloud 17/25
Named London or the UK 12/15
Named Revolut 14/19
Named DPD UK 7/11
Two and a half hours of silence
Kagi posted "europe-west2 outage" at 06:06, two minutes before the start Google later gave. 16 pages had posted by 07:00, 36 by 08:00, and 50 by the time Google's first update came at 08:35, or 01:35 in California. That update spoke of "elevated errors and intermittent network failures". The power event was only named at 11:12.
The first to say it ahead of Google
06:00 08:35 Google posts
How it spread
How it spread 70 status pages
In its own words By region or press reports
Issue
Major outage
Critical
- 25 named Google Cloud. Databases and queues other companies run on, like Elastic Cloud, Confluent Cloud, Redis, Aiven and Neo4j Aura, and products built on top, like Kagi and Atlan.
- 15 named London or the UK. Kinsta's and Cloudways' London data centres, Bloomreach's UK region, Akeneo's London customers.
- 19 named Revolut, which news reports put among the companies hit: open banking in four countries through Tink, card payments through Curve, even the UK government's GOV.UK Pay. Some went a step further: Mollie and orderBilly named Wero, which named Revolut.
- 11 named DPD UK, also reported hit. ShipStation, Shippo, Veeqo and eight more couldn't create its labels.
A problem at the same time is a clue, never proof. But for these companies' customers, the outage had a bank's or a parcel carrier's name on it, not Google's.
What they called it
What they called it of 70 status pages
Only 21 of the 70 said "Google" or "GCP". 30 named a bank or a parcel carrier instead. If you watch your cloud's status page alone, most of this outage looked like someone else's.
Google closed at 14:00. Many didn't
19 of 25 outlasted Google's incident
UTC 06:0009:0012:0015:0018:0021:00
Google closed its incident at 14:00. Of the 58 pages that finished that day, 43 finished later: Redis at 17:24, Sourcegraph at 17:44, Elastic Cloud at 20:11. And 12 were still reporting a problem when the day ended, 8 of them Revolut's partners.
What it shows
Your vendors' status pages can know before your cloud's does, here by more than two hours. The outage that reaches you may not say "Google" at all; this time it more often said "Revolut" or "DPD". And the cloud's all clear isn't yours: clusters restart, queues catch up, and partners take their time.
Every one of the 70 is on the day's outage page, with what it said.
We counted a status page when its own words on 20 October named Google Cloud, a London or UK region, Revolut or DPD UK. Pages with problems in the same hours that didn't say why are left out. Times are UTC, when each page first said it.