Scheduled refresh Failure help

[ScottKC]

Hi and thanks,

Our PBIS environment is a Premium Capacity P1. We have around 40 semantic models that most refresh each morning from 5am-8am, but a couple that also refresh hourly during business hours.

Yesterday we experienced an odd issue with a semantic model on an hourly refresh... the model didn't appear to have all the expected data (despite refresh jobs showing completed) and then at the next hourly schedule the expected data come in to the semantic model. Delving further into the schedule refresh transactions I saw the following "failures" that ran all night (and crossed the next morning's hourly refresh schedules)... but those other refresh schedules were fine. These are the hourly schedules (in descending time order) back to the previous day:

BTW schedules run at intially 5:30am and then hourly between 8am and 6pm.

Using the Fabric Capacity Metrics Preview App, we did have a capacity spike (for 20 minutes) around 8:20pm the night before, but things seemed fine afterwards that night and into the next morning (until I noticed the above failures which started outside normal scheduled hours: 10:30pm and 0:30am):

I have no idea what has gone on here or why. Can anyone enlighten me? How can I identify the cause and prevent this in the future?

Thanks

Scott

[aj1973]

Hi @ScottKC 

If this event happened once then no need to worry as it was an internal error that caused the failure but if it happens continually then the semantic might need a tune up or scale up the capacity.

[ScottKC]

Thanks for the reply. However, this doesn't explain why the failed jobs started at very odd times outside of their hourly schedule. I.E. the last job on the 18/12/2023 had the job starting at "18/12/2023 18:00:09" and ended at "18/12/2023 18:10:03".

The next couple of jobs the next morning started and finished as expected, but then something started at "18/12/2023 22:30:08" and is associated with the 9am hourlry scheduled on the 19th. I have no idea what or why.