Notifications tab in Azure Portal returns misleading and confusing information.

[asos-martinsmith]

Severity

P3 - Medium (Minor functionality affected)

Describe the Bug with repro steps

The "Notifications" tab in the Azure portal can return misleading results.

We wasted time looking into why we kept seeing messages about an issue that we were confident had been fixed (related to invalid tracked properties) but the hitcount and "last occurred" kept increasing despite this.

After digging around in the underlying diagnostic events table storage I surmise that the current behaviour is likely filtering on timestamp within the last 24 hours, aggregating by ErrorCode and returning something like MAX(Timestamp), SUM(HitCount) and either ANY(Message) or maybe even the first (oldest) message in the group.

So in reality we were getting a message logged about a different issue but this was not shown to us in the Notifications tab.

I think that a fix needed here is to either show an entry for each ErrorCode/Message combination in the Notifications tab or, if it is still to be limited to one entry per error code, only show the data relating to the latest Message for that ErrorCode (and not include different rows with different Messages to that latest Message in the hit count).

As a workaround "Diagnose and solve problems" -> "Availability and Performance" -> "Logic App Down or Reporting Errors" and then expanding the "Workflows Validation Summary" does show the latest occurrence for each individual message.

What type of Logic App Is this happening in?

Standard (Portal)

Screenshots or Videos

(In reality the underlying table storage doesn't have any entry with this message since 07:21 this morning but the UI gives the impression the problem was still occurring at 12:28)

As the action complained about in the error message above had had all of its tracked properties removed and we were still seeing the same error message appearing - with timestamps much later than this removal it gave us the impression that the logic app was somehow still loading an old version of the workflow and led us to investigate around that - rather than just telling us that there was some other validation error that still needed fixing.