Published on November 1, 2026 · The mitza.es team
When a cloud service fails, the problem is rarely “ours”: the outage comes from outside, without warning, and can't be fixed. But the company does decide whether that day it sells, invoices and serves its customers, or whether it sits waiting and staring at a screen. This guide starts from three real outages in autumn 2025, explains what they teach and proposes a continuity plan that a small business can put together on a small budget.
Three outages in less than a month
- AWS, October 20, 2025 (US-EAST-1 region): a problem in the Domain Name System (DNS) of the DynamoDB database service, which according to AWS's analysis originated in a latent race condition in its automation, left an empty DNS record and spread to other services. It affected applications as diverse as Slack, Atlassian, Snapchat or Roblox, for several hours.
- Microsoft Azure, October 29, 2025: an unintended configuration change in Azure Front Door, the service that distributes traffic for many applications, left nodes not loading correctly. It affected Microsoft 365, its admin center and services such as Xbox or Minecraft. Microsoft blocked further configuration changes and rolled back to the last valid configuration. Sources put the incident at roughly eight or nine hours, with slight differences in the times given.
- Cloudflare, November 18, 2025: a change in the permissions of a database doubled the size of a configuration file in the bot management system; once it exceeded an internal limit, the proxy servers failed. It left platforms such as X, ChatGPT, Spotify or Canva, among others, unresponsive for about three hours. Cloudflare clarified that it was not an attack.
In all three cases the origin was an internal error or a configuration change, not a hack. And in all three the lesson is the same: you depend on very reliable providers, but not infallible ones. We have also seen similar effects with specific tools, such as the ChatGPT outages of August 2026.
What an hour of downtime really costs
Before designing anything, work out what it costs for each service to be down for an hour: sales that don't happen, orders that can't be processed, invoices that don't go out, people being paid without being able to work and customers who go elsewhere. It doesn't need to be precise: an order of magnitude is enough to decide where a plan B is worth investing in and where it isn't.
Step 1: the dependency map
Make a list of everything your business uses in the cloud or over the internet, and for each item note who hosts it and what would happen if it failed for four hours:
- Email and calendar.
- Invoicing software, ERP or CRM.
- Point-of-sale terminal and payment gateway.
- Website and online store (and the service that serves its images or its DNS).
- Shared files and documents.
- Phone and messaging (WhatsApp, Teams, Meet…).
- AI tools, time tracking, payroll, etc.
Often the finding is uncomfortable: several services depend, without you knowing it, on the same provider behind the scenes.
Step 2: decide what is critical and define a plan B for each service
- Email: if your provider goes down, how does the customer reach you? Have an alternative channel (phone, business WhatsApp) and let people know about it. Consider browser access as well as the desktop program.
- Sales and payments: is there a way to take payments differently (a card terminal on another connection, Bizum, bank transfer)? Can you write orders down by hand and enter them afterwards?
- Website and store: have a static notice page ready that you can publish somewhere else, and know who controls your DNS. We explain that part in domain and hosting.
- Files: keep a synced copy of your critical documents locally or on a NAS so you can open them without internet.
- Internal communication: agree in advance how the team is notified when the usual tool fails.
Step 3: independent copies of your data
A provider outage is temporary, but if the only copy you have lives inside that provider, your room to maneuver is zero. An independent, tested copy, following the 3-2-1 backup rule, is the foundation of any continuity plan. If you use cloud storage, remember that syncing is not backing up.
Step 4: connection and electricity
Not every outage comes from the cloud. A backup internet line (4G/5G) costs little and prevents many stoppages; a UPS protects equipment and drives from power cuts. Also check what happens if the fiber line goes down: do the card terminal, the backup router and the mobile phones keep working?
Step 5: communication during an outage
- Prepare a note for customers in advance (“we are experiencing an incident with a provider; we are available via…”).
- Decide who sends it and through which channel (social media, phone, a sign, an automatic message).
- Check your providers' official status pages before assuming the fault is yours.
- Note what has been lost (orders, hours) so you can claim service credits or adjust the plan.
Contracts and guarantees: what they really promise
The big providers publish service level agreements (SLAs), but the usual compensation is service credits, not compensation for your losses. Read what they guarantee, from when they count an outage and how the credit is requested. And if you depend on a single service to invoice, ask yourself whether it's worth paying for redundancy or keeping a basic alternative.
Step 6: test the plan once a year
A plan that has never been rehearsed tends to fail the day it is needed. Spend one hour a year simulating an outage: switch off the main service “on paper” and check that everyone knows what to do, that backups restore and that the phones and alternative access work.
Common mistakes
- Believing that “the cloud” already includes business continuity.
- Having email, website and DNS with the same provider and no alternative access.
- Not knowing who holds the domain keys or how the DNS is changed.
- Never having tested a backup restore.
- Communicating late or without transparency with customers.
- Investing in redundancy for a minor service and leaving the critical one without a plan B.
How this fits with your cloud strategy
None of this is an argument against the cloud: for most small businesses it is still more reliable and cheaper than their own server. It is an argument for using it with your eyes open. If you want the full picture of models, advantages and risks, read our guide to cloud computing for small businesses.
How we can help at Mitza
We help you build the dependency map, define plan B for what is critical and get the backups and alternative access ready. If you're in southern Madrid, see our IT maintenance service or write to us from the contact page.