Erratic Model Processing Times - Only Source is Gen1 Dataflows

[MartinMason]

We have a Power BI semantic model that is composed of several datamarts, all sourced from a Gen1 Dataflow table. The processing time is usually 45m-1h. However, we're now seeing very erratic processing times, sometimes so long that we're needing to kill processing after several hours. It just seems to hang. The only data source is Gen1 dataflow tables, all in the same workspace as the model. Is this potentially a Power BI service issue or is there something that can be looked at to optimize? The size of the model is <4GB and is composed of 212 tables.

[MartinMason]

Shared. Premium Per User. In the cursed Central region. In process of migrating over to Fabric and Direct Lake. But for now, it is what it is. Architecture is medallion like. Set of dataflows for the bronze pulling from on-prem SQL, Onedrive, Sharepoint, Azure SQL, APIs, Gen2 storage, etc. Set of dataflows to transform bronze to "silver/gold". Semantic model built on top of "silver/gold" dataflows. Been stable for a couple of years and only over the last 8 days has the processing become as crazy as what it has been lately. My coworker blames it on me as I went on vacation last Friday. She thinks the semantic model is like a cat and everytime I leave it, it gets angry and seeks revenge.

[lbendlin]

Shared capacity means exactly that. You are at the mercy of all the other capacity users playing nice.

Your coworker may be onto something.

[lbendlin]

"composed of"  is a bit of a strong word.  "sourced from"  might be more appropriate.

What's the business reason behind the datamart? What are the things you cannot do with the dataflows or with the semantic model directly?

[MartinMason]

Here's a bit more of an illustration of the erratic processing times we've been seeing over the last week or so.

The only data source for the semantic model is Power Platform dataflows so it's not being blocked by an open transaction. That's not in the equation. Somthing very, very odd is going on. Just can't figure out what.

Trying to start a Profiler trace to monitor processing but get this error when trying to start one.

[lbendlin]

What's the capacity SKU? or is this on shared capacity?