Fabric Is Just Plain Unreliable, and Microsoft’s Hiding It.
Update 2025/07/21: Microsoft heard the complaints, and fixed the reliability added a better status page. Baby steps.
Last week, Microsoft Fabric went down yet again for hours on multiple continents.
Oh, you didn’t hear about it? Let’s talk about why.
First, Fabric’s status page is fabricated bullshit. The link https://aka.ms/fabricsupport takes you to a localized status page that almost always shows all green checkmarks – even when the service is on fire. During last month’s 12+hour overnight outage, people were screaming on Reddit overnight that things were down, but the status dashboard was showing all green. When Microsoft employees woke up, they asked if people were still having problems – and then eventually got around to updating the status page to reflect the outage when it was clear that things were really borked.
Redditors have resorted to relying on reporting Fabric outages to Statusgator, who then tracks the time gap between a burst of user outage reports, to the time Microsoft actually updates their status page – and it ain’t pretty:
Second, the post-mortems are just as fabricated. After last month’s outage, the team posted on Reddit, and opened with this whopper:
Fabric/Power BI is deployed in 58+ regions worldwide and serve approximately 400,000 organizations, and 30 million+ business users every month. This outage impacted 4 regions in Europe and the US for about 4 hours.
See what they did there? They used big giant numbers to talk about the subscriber base, and then switched units of measure to talk about the affected population (just “the US”.) That’s like saying, “We served over 30 billion hamburgers last month, but unfortunately, just 1 country (the US) came down with food poisoning.” Gimme a break. Furthermore, the 4-hour thing is just wildly incorrect, as evidenced by the people screaming on Reddit overnight and into the morning.
The combination of factors that triggered this issue did not occur until we hit specific regions and usage patterns. This was caught at that point through automated alerting, and our incident management team initiated a rollback.
Specific regions like, uh, Europe and the United States. You know, small places. Villages, practically.
I absolutely love the second sentence as a world-class example of fabrication. Microsoft is accidentally admitting that their own internal alerting showed that Fabric was broken – but not their external alerting, aka their status page. They’re accidentally showing their cards that the status dashboard just doesn’t show the truth.
Next, Microsoft hides the Fabric outage history as quickly as they can. The status dashboard has no list of recent outages. I feel genuinely sorry for Fabric admins who struggle troubleshooting failed Fabric processes that were supposed to run overnight. They think it’s their own problem, not realizing that there was an overnight outage that Microsoft has simply swept under the rug as quickly as possible. The admin checks the status page, sees nothing, and continues troubleshooting, thinking it’s their problem.
Contrast this with the overall Azure status page, which has a prominent link to Azure status history, publicly calling out major outages and their post-mortems. Microsoft knows how to do this – but the Fabric team ain’t doin’ it.
I don’t understand why the Fabric team is so secretive about the outages.
It’s not like Microsoft Fabric even has a service level agreement.
It’s not like they’re giving refunds when your data is gone for hours at a time.
Oh you didn’t realize that?
That brings me to the only reason I can think of that someone would recommend Microsoft Fabric as a critical part of a company’s infrastructure today: ignorance. That’s where the blog post comes in, dear reader – I don’t want you to be ignorant.
28 Comments. Leave new
Honestly makes me hesitant to take any job in a Microsoft environment under fear of being dumped with Fabric.
Don’t worry there are many Microsoft environments who only use Power BI for reporting/dashboarding and SQL Server / Data Factory for ETL and DWH.
… but PowerBi licences strategy is forcing user to go on Fabric…
I know about SSAS->PowerBI Premium->MS Fabric transitions. But I don’t follow; how is PowerBI client licensing strategy forcing users to go onto MS Fabric and/or OneLake?
I’m a data engineer, so any low code environment is not really for me. It’s still Databricks or bust, really
[…] ???? […]
Crazy, you wouldn’t accept that kind of response from an internal IT team.
If you won’t accept it internally there is no reason to accept it from a vendor hosting services for you. Service meant to replace on premise systems. The minimum should be what you would expect your internal IT teams to provide in response to issues and post mortems.
Hi Brent.
Love the way you put things! Made me chuckle!
Don’t worry – Microsoft fired the coders and AI is now doing the job.
Nothing can go wrong, go wrong, go wrong….
I’m glad I’m not the only person speaking about these kinds of shenanigans from Microsoft. Respect.
Love how honest real BI developers can be on blogs.
Pity LinkedIn is still just a massive big biz brown nosing spectacle.
[…] Brent Ozar points out some major issues: […]
I tried to find News about this outage and couldn’t. The only references to it I found were Brent’s blog and the reddit thread. I thought something like this would get more coverage. Maybe my google-fu is just not up to par?! Did anyone else find any coverage of this outage?
I don’t like the idea that this goes unreported…..that’s kind of the only way to hold companies accountable.
BIgQuery is never down! #justsaying
Nice write up, this is where marketing and PR takes over at Microsoft instead of actually informing the facts.
This is typical in the I.T. industry today, its not just Microsoft.
Pretty sure all those status pages aren’t automatic and need approval from some upper management level to post anything negative!
Ahhhh…reminiscing the good old days when you controlled the infrastructure/platform.
A bit about semantics: the message of MS says „four regions in Europe and the US“ meaning those four regions are located in both continents, Europe and the US.
Not four regions on Europe plus the whole of the US. Other than that, this fact spinning really is a shame.
MS botched implementation of Fabric and lack of investment into SQL Server has turned me off of MS products at large. Wherever possible I steer my company away. I view SQL Server as the old reliable OLTP (which it’s great at), could care less about the new Azure features they continue to tack on. I’ll never use them and suspect most people won’t. Ditto Kendra Little’s recent article about SQL Server. Give us some real features that we actually want. Stop pushing everyone onto this Fabric crap that does not work. It might help revenue in the short-term but long-term will have a negative impact since developers have lost trust. They’re riding the marketing sales pitch wave right now, reality will eventually hit the numbers. As other commenters have mentioned, there’s a growing talent pool that won’t go anywhere near a job that mentions Fabric.
I loved reading your comments. Would you be able and willing to post Kendra Little’s article or send it to me?
They’re probably talking about this article:
https://kendralittle.com/2025/03/10/what-the-decline-of-sql-server-quality-means-developers-dbas/
Yes, thank you for this. Have been saying it since my company started pushing it heavily..:
Hi Brent,
I can not agree more with you, I was one of poor Fabric admin guys that suffered the issue last week… (Western Europe Region) Suddenly all the platform started to do strange things. It’s not the first time this happens (I mean, outage not communicated) and the pattern is always the same. Problems remains hidden or minimized.
See Last year problem when I announced on X a similar scenario
https://x.com/Alf_BI/status/1808401650060501018
and no one else in MS reported it.
At that time I discover alternatives as https://statusgator.com/services/microsoft-fabric but as you said Fabric Team has a serious problem of transparency.
Hi Brent, this was a really interesting read as we have a customer trying out Fabric at the moment and we’re going to raise it as a concern.
The lack of SLA surprised me so I did some digging. It does exist here (as long as you haven’t let it default to 2023 ?) :
https://www.microsoft.com/licensing/docs/view/Service-Level-Agreements-SLA-for-Online-Services?lang=1&year=2025
But to be fair it’s pretty noddy, just mentioning stuff like being able to “Open and view a Notebook (or warehouse) in the service”. Nothing about actually being to run/schedule a notebook, or connect to a warehouse with other tools like SSMS.
That blows my mind.
Open a notebook… open a warehouse… but not actually run anything, that’s the SLA.
I absolutely love Fabric
-signed
David Blaine
[…] been pretty vocal here on the blog and on social media about the reliability problems with Microsoft Fabric. Today, I’ve got good news: Microsoft released a new Fabric status page and a known issues […]
Did read somewhere or have I made this up in my mind that Microsoft committed to improving the failover of Fabric from READ-ONLY to WRITE?
If I haven’t made this up in my mind, can someone please paste the link to the Microsoft article..?
Your best bet for research like that would be to create a post at a Q&A site like SQLServerCentral.com, DBA.StackExchange.com, or Reddit.
Thanks Brent.
I found it from Microsoft’s Post Implementation Review report…
We (Microsoft) will improve Fabric & Power BI failover capabilities to allow read/write when in failover mode. Tentative date = July 2026