· · ·
A Reality Check on Cloud Reliability
The Cloud isn't as infallible as we'd like to think. Test your disaster recovery (DR) and business continuity planning (BCP).
On 29 October 2025, Azure Front Door had a nightmare. An "inadvertent tenant configuration change" (Microsoft for "we pushed a bad config") turned into an 8+ hour outage for organisations worldwide. Microsoft's preliminary Post Incident Review has the detail.
Even the big cloud providers have spectacularly bad days. Your DR plan has to assume that.
Microsoft's "just turn it off" advice
When it broke, the suggestion was essentially: bypass the CDN. That sounds tidy until you list what you lose:
- No WAF or DDoS protection if Front Door was your edge security
- Exposed origins you thought were tucked behind the CDN
- No caching - origins take the full hit and melt
- TLS headaches - especially with certificate pinning
- Private endpoints - for some designs there is simply no alternate route
Bypass is not a DR plan. It is a different outage with extra steps.
The SLA fantasy
Azure Front Door advertises 99.99% - under an hour of downtime a year. This incident alone burned through that for many customers, several times over.
An SLA is a commercial number. It is not a substitute for business continuity. If one bad day can spend a decade of "allowed" downtime, you already know what your architecture must tolerate.
Map what dies when your CDN dies. Keep recovery paths that do not depend on the broken component. Test restores before the next "inadvertent" change - not during it.
Find this on LinkedIn and give it a thumbs up: A Reality Check on Cloud Reliability
Related
- Mitigating Single Points of Failure for HA, DR, and BCP strategies
- Security vs. Compliance for why meeting SLAs is not the same as being secure