pbivisuals.powerbi.com server failure
We're using PBI Embedded with some certified custom visuals including Plotly. These custom visuals have stopped displaying recently and we've traced the problem to simple GET calls on pbivisuals.powerbi.com returning data so incredibly slowly that the visual times out. Occasionally that server will return the requested json definition (all 38MB of it) in 150ms. But the vast majority of times it will respond at less than 10kb per second (or not at all) and the request will ultimately time out. This is very easy to replicate in any browser by trying to access the Plotly visual definition directly: https://pbivisuals.powerbi.com/plotlyJSVisualEEDDFAAB4619388784AE7972BC162ECDFAAB46196.2.1.0.0.cc6d7a5f02f9a7ef6a04543a2a66140847cec0ac298cbe8d7eb853fead875102.pbiviz.json We've tested fetching this via VPN across multiple global locations, via different browsers on different machines etc and is is almost always paralyzingly slow. Is anyone able to suggest how to get Microsoft to look at this? Just to reiterate, our production reporting platform is effectively down atm because of this server-side issue. Thanks for any assistance you can offer!
Can we improve relationship arrow rendering in the semantic model view?
There are multiple issues with the way relationship arrows are rendered, and this image displays a few of them: (Note: I am using the web interface, via Edge.) Is there anything I can do to combat this confusing display pattern? An arrow that points ambigiously is worse than no arrow at all. If anyone in power is reading: please either give us more control over how these are rendered or make them behave logically. Example: in the first and last relationships shown above, I know that the arrow has rotated because the connecting lines are not perfectly aligned, but these lines should automatically align vertically as the others have and the arrow should only rotate if there are visible horizontal connectors extending out its sides.
Unhandled error from powerbi-client embed event listener is surfaced as global browser Script error
Hi Power BI team, we believe we found a bug/limitation in powerbi-client event handling for embedded reports. When a report is embedded in an iframe (cross-origin), an exception thrown in an embed event listener appears to escape as a global browser error (window.onerror / Script error) instead of being contained in the SDK event pipeline. Environment powerbi-client: 2.23.1 Angular: 21.2.x Browser: Chrome (prod build) Embed host and iframe are cross-origin Observed behavior powerbi-client receives a postMessage from iframe, routes it to event handling, and dispatches embed events. If one listener throws during this flow, the app receives a global window error (generic Script error / event-wrapped error), which can trigger app-level global error handling/UI alerts. The global error lacks actionable context and looks like app failure, even though it is iframe/event-listener noise. Expected behavior Exceptions from embed event listeners should be safely handled/isolated by SDK (or surfaced only through SDK error channel), not leaked as global browser errors. At minimum, provide reliable error metadata so host apps can distinguish SDK listener failures from real app errors. Workaround on our side We suppress non-actionable global errors in our Angular ErrorHandler when: message contains Script error, or wrapped error has cause instanceof Event This avoids user noise, but ideally the SDK should not leak these as global errors. Repro outline Embed Power BI report with powerbi-client. Register at least one embed event listener (e.g. loaded, rendered, error). Throw an exception inside that listener. Observe global window error event and host app global error handler invocation.
"When a new row appears" report alert not firing on a Direct Lake (OneLake) model
I've set up a Fabric Activator alert on a Power BI report table using the "When a new row appears" condition, with a unique row identifier column, notifying via Teams. The visual is a table on a report page, and the semantic model is Direct Lake on OneLake. The problem: I added a new test record (a new unique ID) to the underlying data. The record shows up correctly in the report table, but I never received the Teams notification. What I've already checked/ruled out: - The alert is bound to the correct visual, and the unique-identifier column is present in that visual. - The new record is visible in the report, so the model is reflecting the change & the data is present. - There are no filters excluding the new record from that visual. - The semantic model has "Keep your Direct Lake data up to date" turned on My Working Theory: Because this is Direct Lake, the model reframes (re-points to the latest Delta/Parquet files) rather than doing a traditional refresh, and it does so automatically as data lands. My suspicion is that the Activator "new row" rule only re-revaluates when it receives a semantic model refresh event - which a reframe does not raise - so the report updates but the rule never runs to detect the new identifier. My questions: Has anyone successfully used the "when a new row appears" report visual alert against a Direct Lake model, and had it fire correctly? Does this alert type require a scheduled/triggered refresh (reframe) on the Direct Lake model to evaluate? if so, does forcing a reframe (refresh API or the Semantic model refresh pipeline activity) after each load reliably trigger it? If the report visual path isn't dependable for batch-loaded Direct Lake data, what pattern are you using instead (e.g. scheduled KQL Queryset > Activator, OneLake events, or a pipeline driven notification)? Any confirmation of the refresh-vs-reframe behavior, or proven design pattern, would be much appreciated. Thank you!
