Contact Sales
Get Started Now

Google Cloud London head: impact on Revolut and architecture of redundancy

24Cevent Sin categorizar Google Cloud London head: impact on Revolut and architecture of redundancy

Google Cloud London: 7 hours of blackout that left Revolut not work in Europe

October 20, 2026, Google Cloud experienced a break critical in your region europe-west2 (London) that lasted approximately 7 hours and 52 minutes (06:08 UTC at 14:00 UTC), caused by a mains power failure in the area of europe-west2-b of the datacenter. More than 30 Google’s Cloud services were impacted, including AlloyDB, BigQuery, Cloud Run, Cloud Filestore, Cloud Firestore, Google Compute Engine, and Cloud Spanner. But the more visible impact was in Revolut, the fintech european leader, who was completely inoperative: users in Spain, the Uk and France could not access the app, check balances, transfer, or use cards. Revolut confirmed the cause: there is a technical failure on servers of Google Cloud. For organizations throughout Latin america that depend on cloud providers, this incident illustrates a truth operational terrifying: when your cloud provider of have a problem, YOU have a problem, and during those 7 hours, no fallback, no contingency, just wait.

What happened exactly in Google Cloud on the 20th of October?

23:08 PDT, October 19 (06:08 UTC on the 20th of October), there was an event of energy in the datacenter europe-west2-b of Google Cloud in London. There was a cascade of components failing. It was not an attack. It was not a software error occurred. Was able to: a failure in the electrical infrastructure of the datacenter.

When you lose power in a datacenter, you lose it ALL: server will not start up, cooling failure, local redundancy does not work. Google Cloud started to investigate, but the damage was already done: tens of thousands of instances in that area were down.

Gravity is remarkable: Europe-west2 is the region cloud that is nearest to european users. Whether you’re building a service fintech in Europe, there is a very high probability that use Europe-west2 for low latency. Revolut uses Google Cloud. HSBC uses Google Cloud. Lloyds uses Google Cloud. When Europe-west2 falls, all fall.

What were the services affected?

In Google Cloud directly:

More than 30 services impacted:

  • AlloyDB: Database-PostgreSQL-compatible
  • BigQuery: Data warehouse and analysis
  • Cloud Run: Serverless compute
  • Cloud Filestore: file Storage
  • Cloud Firestore: NoSQL Database
  • Google Compute Engine: virtual Machines
  • Cloud Spanner: distributed Database
  • And 20+ additional services

In Revolut (the most visible):

For users of Revolut in Spain, UK, France:

  • App completely blocked with the message ‘Something went wrong’
  • Lack of access to accounts
  • Transfers arrested
  • Declined cards in shops
  • Withdrawals at ATM failing
  • Balance is not visible

For the internal operations of Revolut:

  • Inability to process payments
  • Systems compliance offline
  • Dashboards operations unreachable

What was the cause of the incident?

Cause confirmed: power Failure in zone europe-west2-b of the datacenter of Google Cloud.

This is critical because it reveals a vulnerability: while Google Cloud has redundancy, the redundancy WITHIN the same datacenter (different areas) does not protect against failures that affect the datacenter full of energy.

Redundancy true requires:

  • Multiple datacenters geographically distant (not only of zones within the same datacenter)
  • Networks of independent energy (not the same electric provider)
  • Automatic Failover to the region’s alternative

Revolut seems to not have that, or if it has, it did not work properly.

How long did it last exactly?

Official duration: 7 hours 52 minutes (06:08 UTC at 14:00 UTC on October 20).

But the reality was more complex:

  • 06:08 UTC: Power event in europe-west2-b
  • 06:10-06:30 UTC: Google detects the problem, begins research
  • 06:30-08:00 UTC: Users notice problems (Revolut app does not load)
  • 08:00-12:00 UTC: Google is working on recovery, services partially restored
  • 12:00-14:00 UTC: progressive Recovery of the majority of services
  • 14:00 UTC: closed Incident
  • 14:00-16:00 UTC+: Some customers continued to report slow or limitations

For end users, were 7-8 hours without access to digital money.

What was the operational impact real?

To Revolut

  • Millions of transactions blocked
  • Frustrated users trying to pay, transfer, withdraw
  • Reputation risk: ‘My money was locked 8 hours’
  • Loss of confidence: ‘do I have to change to traditional bank?’
  • Cost regulatory: Reports to authorities, explanations, possible fines
  • Opportunity cost: Traders without access to crypto conversions, users without international transfers

Google Cloud

  • Damage to reputation: ‘Google Cloud is not reliable’
  • Customers evaluating alternatives: Azure, AWS
  • Forensic analysis: internal Investigation as to why it failed redundancy
  • SLA violations: Compensation to customers with SLA
  • Lost confidence: Startups fintech reassessing whether to use Google Cloud

For users

  • 7 hours without access to money
  • Transactions that were not completed
  • Travel interrupted (card is not working)
  • Financial emergencies blocked

What can we learn?

This incident teaches critical lessons about architecture, redundancy and management of incidents:

1. Redundancy within the same datacenter is NOT enough

Revolut (and Google Cloud) have ‘zones’ within europe-west2 for redundancy. But if the problem is the datacenter full (power, cooling, network), zones collapse together.

Lesson: Redundancy true requires multiple regions in geographies DIFFERENT.

2. Detect problems with the provider, it is your responsibility to

When Google Cloud fell, Revolut should be detected instantly that their requests to the Google Cloud were being rejected. But it seems that it took to detect and communicate to users.

Lesson: automatic Monitoring of dependencies outside is critical.

3. Automatic Failover to the region’s alternative can save the day

If Revolut had replicated data in europe-west1 or us-central1, would have been able to do automatic failover in minutes, not 7 hours.

Lesson: For fintech, data replication and automatic failover is a requirement, not a luxury.

4. Transparent communication during outage is critical

Revolut took time to communicate. Users learned on Twitter that he had a problem, not for Revolut. This increased panic.

Lesson: automatic Notification to users about the status of incident should be immediate.

5. The events of power in data centers are predictable but manageable

Google Cloud should have generators, UPS, multiple power supplies are independent. If Revolut depended entirely on Google Cloud without redundancy, that was the decision of architecture.

Lesson: For critical operations, single cloud provider is an unacceptable risk.

How is it related to the automatic management of incidents?

This incident illustrates a common problem in fintech: dependence on suppliers without visibility or preparation for failure.

Management manual:

  1. Google Cloud power event (06:08 UTC)
  2. Users will notice the app does not load (06:15)
  3. Support get reports via email/chat (06:20)
  4. Someone investigates, he discovers that Google Cloud is down (06:30)
  5. Manager scale escalation of the crisis (06:45)
  6. Communication to users via Twitter (07:00)
  7. Wait for Google Cloud recovers (up to 14:00)
  8. Validate that Revolut works again (14:30)

With management intelligent auto:

  1. Google Cloud power event (06:08 UTC)
  2. System Revolut detects which requests to Google Cloud fail (06:09)
  3. Automatic correlation: ‘Google Cloud eu-west2 is down’ (06:10)
  4. 24Brains runs validation: confirm that a region alt. is OK (06:11)
  5. Automatic Failover to the region’s alternative (06:12)
  6. Users are notified of switch region alternative (06:13)
  7. Users continuán operate with greater latency but full functionality
  8. While Google Cloud is retrieved in parallel

Difference: 8 hours without service vs. 5 minutes with a small degradation.

How can you help 24Cevent?

Although 24Cevent is incident management, IT, the principles apply directly to services that depend on cloud providers:

1. Automatic detection of failures suppliers

24Cevent can continuously monitor the availability of critical dependencies (Google Cloud, AWS, etc) and to detect degradation instantly.

Instead of waiting for user reports, your system would know the 10 seconds that Google Cloud europe-west2 is failing.

2. Correlation smart symptoms complex

When multiple services fail (Revolut, HSBC, Lloyds all simultaneously), correlation smart identify instantly the common cause: ‘Google Cloud provider issue’, not ‘Revolut have database problem’ + ‘HSBC has API problem’ + ‘Lloyds have auth problem’.

3. 24Brains run automatic diagnosis

In the time that it detects degradation, 24Brains can automatically:

  • Validate connectivity to the Google Cloud
  • Review the status page of the Google Cloud
  • Run health checks on region alternative
  • Compile diagnosis: ‘Google Cloud eu-west2 is down, us-central1 is operational’

All in 30 seconds, without human intervention.

4. Workflows automatic failover

Configure workflows pre-defined:

  1. If it detects failure of Google Cloud eu-west2
  2. Validates that region alternative is OK
  3. Running automatic failover (redirected traffic)
  4. Notifies users: ‘Using fallback region, expect 50ms higher latency’
  5. Scale to engineers if fallback also fails

5. Notification multichannel automatic

Instead of waiting for someone to tweet about the issue, 24Cevent notified automatically:

  • Users: Status page updated automatically
  • Support: Slack notification with a technical background
  • Management: Email executive with estimated impact
  • Regulators (if required): Report of the disruption of service

6. Complete traceability for analysis regulatory

From the moment of failure up to resolution:

  • 06:08 UTC – Google Cloud power event (external)
  • 06:09 UTC – Revolut detects API failures to GCP
  • 06:10 UTC – Incident correlated as ‘GCP provider issue’
  • 06:11 UTC – 24Brains diagnostics completed
  • 06:12 UTC – Failover to us-central1 initiated
  • 06:13 UTC – Users notified
  • 06:14 UTC – Monitoring confirms failover is successful
  • 14:00 UTC – GCP recovery confirmed
  • 14:15 UTC – Failback to eu-west2 completed

For regulators and auditos, this is crucial: proof that you detected fast and you answered systematically.

7. Free Plan for startups fintech

For startups that build on Google Cloud, 24Cevent Free plan allows you to:

  • Fault detection provider
  • Automatic notifications
  • Workflows basic
  • Up to X contacts

Frequently asked questions

How much money you lost Revolut in those 7 hours?
Conservatively, millions. If Revolut process 1000 transactions per minute on average, 7 hours = 420,000 transactions blocked. $5 average commission = $2.1 M in transactions. In addition, loss of confidence is unquantifiable.

Why Revolut does not have failover to another region?
Probably because of cost (replication of data is expensive) and complexity (failover is difficult to implement correctly). But for fintech, is no excuse.

What should have SLA compensation?
Yes. Google Cloud offers SLA of 99.95% in multi-region. This allows you to ~22 minutes downtime/month. 7 hours violates greatly that SLA.

What is an ‘automatic failover’?
System detects failure of datacenter A, validates that datacenter B is OK, and automatically redirects traffic from users to B without manual intervention.

How do you prevent 24Cevent this?
Does not prevent failure of Google Cloud (out of control). But detects and responds automatically in minutes, allowing failover manual or automatic, reducing to 7 hours to ~15 minutes impact to users.

Sources

LinkedIn
X
Reddit
Facebook
Threads
WhatsApp

Recent posts