Three Backups, One Point of Failure: What a 4 a.m. Outage Taught Me Before an SAP Cutover

A Telekom fibre and mobile outage, unannounced maintenance in the heart of Budapest – and why it's exactly the same lesson as an SAP cutover weekend

Event: Telekom fibre and mobile internet outage

When: Thursday, 15 October 2026, 04:00

Affected: fibre internet, Wi-Fi calling, mobile internet – and customer authentication

Telekom's explanation: maintenance that took longer than expected – without advance notice

Last week I was asked to support a weekend cutover for an SAP brownfield migration. If you've done one, you know the mindset. You plan for what breaks. Every critical step has a fallback, every fallback has an owner, and you hunt down single points of failure before they hunt you down.

Then, on Thursday 15 October, at 4 a.m., my own home infrastructure taught me the same lesson. Free of charge, as a hands-on training session, on my terrace, in the dark.

The Setup – Everything Fine on Paper

Since February I've lived in a new residential complex in District IX (Ferencváros), right in the middle of Budapest. On paper I'm well covered:

  • Telekom fibre with an optical router – my primary connection.
  • A USB mobile-internet stick with a Telekom SIM – my backup.
  • Both Telekom and One are present in the complex, so in theory there is a choice.

In practice, there is almost no Telekom mobile signal inside the building. My phone only works through Wi-Fi calling. My USB backup stick only works where there is mobile signal – which is to say, not in my apartment.

I'm not alone. The complex has more than 240 neighbours, and the same coverage problem keeps coming up among us. When the topic of the underground garage came up, Telekom's answer was a solution there – for an additional monthly fee. Coverage in the centre of the capital, it turns out, can be a paid add-on.

The Night It All Lined Up

Thursday, 15 October – 04:00
The fibre goes down
A nuisance at first. Until I realised what it took with it.
A few minutes later
The domino effect
No fibre → no Wi-Fi.
No Wi-Fi → no Wi-Fi calling → no phone.
The USB backup stick → useless, because there's no Telekom mobile signal in the flat.
Before dawn, 7 °C
Out to the terrace – one bar of 5G
The only place with signal was outside: one bar of 5G. I could make calls – but even there, mobile internet didn't work. So the failure didn't only hit the fixed-line fibre service; it hit mobile too.
Start of the call
Vanda and the customer number
Vanda, the virtual assistant, asked for my "MT" customer number, which I was supposed to look up in the app. The app doesn't work without internet – and I had none, neither at home nor on mobile. Vanda couldn't authenticate me.
About 2–3 minutes later
An automated apology
An SMS arrives: Telekom is sorry that my conversation with Vanda, the virtual assistant, had been unsuccessful.
Between 04:00 and 05:00
Three calls with Telekom
After waiting, I managed to talk to Telekom three times. The feedback: they were performing maintenance that was taking longer than expected.
Beforehand
A notice? There wasn't one.
I received no advance notice of the maintenance whatsoever.
Three channels, one dependency. Fibre, mobile internet, Wi-Fi calling – and even customer authentication depended on the same network that had just gone down. Engineers call it a common-mode failure. I called it a long morning.

On the Terrace with Vanda

So there I was, on my terrace, in the dark, at 7 °C, with a single bar of 5G, calling customer service. First I reached Vanda, Telekom's virtual assistant. Vanda asked for my "MT" customer number – and told me to open the app to find it. But the app doesn't work without internet, and I had none: not at home, not on mobile. Vanda couldn't authenticate me. I kept trying.

About two or three minutes later, an SMS arrived. Telekom was sorry that my conversation with Vanda, the virtual assistant, had been unsuccessful.

The failure worked perfectly. Even the apology was automated.

Between 4 and 5 a.m., after waiting, I managed to talk to Telekom three times. The answer was the same each time: they were performing maintenance that was taking longer than expected.

On paper the process worked: there was a contact, a response and an apology. In reality I was exactly where I started – no internet, no phone, on a cold terrace. These are the moments that cost you not only time, but nerves – and probably a few more gray hairs.

Maintenance – Without Notice

I received no advance notice of the maintenance. If it was planned, why weren't customers told beforehand? And if it took longer than expected – while taking down both fibre and mobile internet – then something probably didn't go to plan. Did it get stuck? Did it fail? Was there a rollback plan?

I don't know. But maintenance that, before dawn and without notice, takes down fixed-line and mobile data at the same time, while the customer can't even authenticate, isn't maintenance. It's an incident.

Not Out of the Blue

The missing coverage isn't news. I filed a formal complaint with Telekom back in March, the emails went on for months, and in the end the property management received an offer: paid coverage for the underground garage only. For the flats, Telekom still recommends Wi-Fi calling – exactly what disappeared together with the internet that morning.

I've written up the full correspondence, how long each step took, and my questions in a separate case study.

What This Has to Do with SAP Cutovers

Everything in that story is a pattern I see in migration projects.

1. Backups that share a dependency with the primary aren't backups

My fibre, my phone and my USB stick all depended on one provider. On a cutover, that's the fallback system on the same network segment, or the rollback that needs the very system that just failed.

2. Automation without a human escape route amplifies failure

A bot that, in the middle of an internet outage, asks you to authenticate through an app that needs the internet, followed by an automated apology, closes the loop on paper and leaves the customer exactly where they started. Hypercare channels need the same thinking: when the automated path fails, who picks up?

3. Known issues become normal

When 240+ households have lived with the same gap since moving in, it stops being an incident and becomes the environment. The garage answer – a fix for a fee – treats a structural gap as an upsell. Projects do this too, with workarounds that quietly become permanent.

4. Test your fallback under real failure conditions

My backup looked fine until the moment I needed it. Dress rehearsals exist for exactly this reason – not to prove the plan works, but to find out where it doesn't.

5. Maintenance is a change too – with notice and a rollback plan

Unannounced maintenance that overruns and takes down both fixed-line and mobile data along the way – in an SAP project, that would be a change with no communication and no working rollback plan. On a cutover, the maintenance window is known in advance, we know who notifies whom, what the go/no-go criteria are, and when we roll back if things don't go to plan.

Practical Tips If You're in a Similar Situation

  • Make your backup truly independent. Different provider, different technology, different physical path. If mobile internet is your backup, choose a provider that demonstrably has signal inside your home – measure it on the spot, in several rooms.
  • Before moving in, test mobile coverage inside the apartment, not just on the street or on a coverage map.
  • Test your backup with a real failure. Unplug the optical router and see what still works: internet, phone, and the ability to report the fault.
  • Plan your fault-reporting path in advance. If your phone relies on Wi-Fi calling, reporting the outage won't work exactly when you need it.
  • Keep your customer number offline. If customer service asks for the customer number from the app, it won't be reachable during an internet outage. Write it down, or save it on your phone offline.
  • Shared problem – shared action. When hundreds of households in one complex are affected by the same gap, report it together and in a documented way. That is what moved things forward in my Digi case as well.
The lesson in one sentence: it doesn't matter how many backups you have, only whether they share a common point. If they do, you really have one system – in several pieces.

Related Case Studies