Fabric has gained features and a better monitoring interface, but there are also concerns about telemetry and billing transparency. How should teams weigh those changes when considering it for production?
Is Microsoft Fabric getting better, or are the operational gaps getting worse?
Answers
The monitoring changes are a good example of the tradeoff. A cleaner interface helps, but it doesn't replace being able to connect a capacity spike to the query that caused it.
There is a concern that the explicit link between query insights and capacity metrics has been lost, with an AI-generated explanation offered instead. For performance investigations, I'd check whether the available telemetry lets your team verify that connection rather than relying on a probable explanation.
Reply to this answer
Billing would be my sticking point. The concern isn't just the price, it's whether you can see how charges were calculated and attribute them to the work that generated them.
Before committing workloads, ask for the current billing details and test them against a representative workload. A cost estimate is much less useful if nobody can explain the bill afterward.
Reply to this answer
I'd judge production readiness on the boring stuff first. Can the team trace costs, manage identity, deploy changes reliably, and schedule jobs the way it needs to?
New integrations are useful, sure, but they don't make those operational gaps disappear. There are concerns about scheduling and Git workflows in particular, so test your actual deployment and scheduling requirements rather than assuming they work like other tools.
Reply to this answer
There's a broader platform choice here too. Fabric may be appealing if its integrations fit your stack, but think about how dependent your workflows would become on it and what moving them elsewhere might involve. That's separate from whether any one new feature is good.
Reply to this answer
Add your answer
Answers are reviewed before publication.
q.493!marblefob:
Also check the cost of the investigation itself if the workflow relies on an AI agent. That can matter when troubleshooting is frequent.