Fabric, Power BI, Power Platform, Data Platform: My session at Fabric Week by ESPC

On Tuesday, May 27th, I had the honor of presenting a session for the Fabric Week at ESPC:

A Day In The Life Of A(n Accidental) Fabric Administrator.

The session was recorded and is already available on YouTube, but unfortunately we ran out of time before I could answer all the questions during the session.

Thanks to ESPC and Seán Colleran I got hold of all the questions asked during the session, and I will provide answers in this blog post where I can.

Q&A

Without further ado, let's go to the questions:

How do you enable Fabric mirroring from the tenant settings?
You don't have to enable mirroring specifically, but you do need to enable the creation of Fabric items in general: Enable Microsoft Fabric for your organization. You also need a workspace with an active Fabric capacity (trial or any F-SKU).
Can we mirror PostgreSQL with that option?
Yes, you can mirror PostgreSQL, but only Azure Database for PostgreSQL flexible server. Keep in mind that currently mirroring doesn't support Azure Database for PostgreSQL flexible server behind an Azure Virtual Network or private networking.
As a Fabric admin, can I scale Fabric Capacity from F64 to F128, for example?
For scaling only, being a Capacity admin should be enough, as explained in scale capacity.
What’s the best way to implement stop/start on Fabric Capacity?
You might be able to, depending on the rights you have in the Azure Portal.
To be specific, you need to have at least contributor rights on the Azure subscription. Specific permissions (to create a capacity) for an RBAC role are explained here.
You can use the Fabric Capacities REST API's to Pause/Resume the capacity. You can call these API's with PowerShell, but also with Azure Data Factory, a Logic App, or from within a Fabric notebook. For the latter, you need a different capacity then the one you want to pause/resume though.
I want to manage capacity units. How can we manage the autoscaling function?
If you mean Autoscale Billing for Spark, that is currently in preview and only controls the spark compute. There is no other option for autoscaling the SKU-size in Fabric.
We have multiple customers on one tenant. Each wants to monitor their own Fabric capacity. What’s the best practice for setting up the Capacity Metrics App?
In the Capacity Metrics workspace, you can create multiple copies of the report and put a report filter on a specific capacity. Then give a user specific permissions to only view that report.
Is there a way to reset tenant settings to “default” or Microsoft-recommended settings?
Unfortunately not. You can check the documentation for the Tenant settings index, some settings mention a default value.
If you are able to set up a totally new (test) tenant in Azure you could check the default values for the Tenant settings in that tenant.
On the permissions level, how do you differentiate between non-production and production environments?
Usually this is done by using workspaces for dev-test-prod. On a workspace level you can easily manage permissions. Within Fabric you can use Deployment Pipelines to transfer items between the dev-test-prod stages.
I’m a user. Our new admin has set up Fabric Capacity (F2 pay-as-you-go). I have permission to create workspaces, but the Fabric Capacity option is greyed out.
To be able to assign workspaces to a capacity you need at least Contributor permissions. The admin of a capacity can assign users on the Capacity settings page in the Admin portal:
Who can we scale up Spark clusters? How can we ensure only admins can do this?
So go from a small Spark pool to a medium Spark pool.
A workspace Admin can change the settings for Spark workloads in the workspace settings:

They can adjust the current pools and create new ones. You can't limit it further, so every workspace Admin can do this.
If you uncheck the last option, Members or lower can't change the settings on a session-level.
What’s the best way to implement a stop/start schedule for Fabric — e.g., only running during the day?
Please see question 4.
For notification activities in a pipeline: I want to use Teams notifications. Do I need an Office license or a specific Teams license?
To actually set up a working Teams activity, you need a Teams license, and a channel or chat to post it to. I was able to add the activity without a Teams license, but I can't select a channel to post to in that way.
Can a workspace admin restrict a tenant admin from accessing a specific workspace?
No. A tenant (Fabric) admin can always go to the Admin portal and gain access to any workspace in the tenant.
What is the best alert mechanism? As a Fabric Admin, I want to know about overloads as soon as possible.
There are a few options, I'll let you decide on the best one to use.
On the Capacity settings page you can set up a notification for X% capacity usage.
Somewhat similar, but more customizable, you can also set up an alert on the Capacity Metrics app with Activator (Real-Time Intelligence) and send a Teams or Outlook message.

Not an alert mechanism per se, but on the capacity settings page, you can also set up Surge Protection, which lets you limit the compute consumed by background jobs. This way you can make sure the capacity doesn't overrun from background jobs.
Is there a way to kill a process in interactive operations?
For most interactive operations you can't, like DAX or XMLA queries. Lakehouse or Warehouse queries can be stopped with the KILL command used in T-SQL.
If I pause a capacity with over-usage, will I see in advance how much I'll have to pay?
No, not out-of-the-box. You could calculate the amount of overage, CU's and thus price, mainly from the Capacity Metrics app. But that's a manual calculation you'd have to do upfront.
Is there a way to connect that monitoring to a running Grafana?
Apparently you can connect a Managed Grafana instance to a KQL database (eventhouse) in Fabric. It's explained in this post.
Do you need to be paying for Purview to use the Fabric Purview Hub reports?
The Purview Hub reports are included as part of the Admin Monitoring workspace, so you don't need to pay extra or for a Purview license.
Is there a way to monitor the count of accesses per user on a report (like in PBI Report Server ExecutionLog3)?
You can get this data from the default Usage Metrics Report, generated when clicking the below option on a report.

You can also get all activity data (from all reports) out of the Activity Log, for example with the REST API - Get Activity Events. Keep in mind though, this has a data retention period of 30 days.
I understood the admin monitoring semantic models were locked — how did you create a custom version?
You can indeed not edit the model itself, but as a Fabric Admin you do have Member access to the Admin Monitoring workspace, and thus Build permissions. So if you connect to the semantic model from Power BI Desktop, you can extend it (create a local model) with your own data. You would also be able to do this with Viewer access and build permissions.
Pro vs. Premium per-user license: What are the three most important reasons to choose Premium per-user over Pro on a Fabric capacity?
It depends? 😉 In most cases, I would say if you already have Fabric, PPU is not necessary anymore. But of course, there are edge cases where it might benefit to have a PPU, I can think of one: Model size limits in low F-SKUs are pretty low (3 GB), in PPU it's 100 GB
Have you seen examples where the 30-day limit for admin monitoring has been extended by updating a custom Lakehouse and running reporting off that?
Yes, you can export the Admin monitoring data to a Lakehouse. The solution I called out in my presentation is FUAM, which doesn't specifically extract the data from the Admin monitoring workspace, but it gets similar data from the rest of the tenant.
If I am not a Fabric admin, how can I get access to admin monitoring?
By default, only users with the Fabric admin role (or higher) will have access. But the Fabric admin can also give other people (Viewer) access to the workspace.
Is there an API to extract user views on each Power BI report page?
Yes, the Activity Events API, see question 19.
The 30 days... is that the same as the Dashboard Usage Metrics?
Yes, right now, the retention is similar in all the Power metrics reports, except for the Capacity Metrics, which holds 14 days of data.