Power BI service Refresh failure

[hoosha_11]

Hi,

I have scheduled a dataset refresh in PBI service every 30 minutes. Currenctly, I'm running it as a test to make sure it operates smoothly. Although all the refreshes are completing, after checking the logs, I noticed that some refreshes failed initially (with the error message: "This operation was canceled because there wasn't enough memory to finish running it"), but succeeded in second try.

The refreshes usually takes 7 minutes to complete, but when it fails, it takes 1':30" to fail:

I have 2 questions:

1) If some operations are canceled due to insufficient memory, why do subsequent refreshes succeed without errors?

2) If a scheduled refresh fails four times in a row, Power BI disables the refresh. Based on this, should I be concerned? I've already implemented incremental refresh and removed any unused columns and tables.

Thanks.

[lbendlin · Accepted solution]

1) You're not the only workload on that capacity. Performance varies based on workload, time of day, seasonality etc

2) You need to distinguish between retries and full failures.  Your example doesn't count as it eventually succeeded. It has to completely fail all four tries  four times before the schedule is disabled.

Having said that - whatever you do to speed up the refresh is absolutely helping. Remember there is a hard limit (5 hours in Power BI service, but can be shorter in the data source) and you want to stay well under that.

[aj1973 · Accepted solution]

Hi @hoosha_11 

1) In a non Premium capacity workspace, this late is a shared capacity (Pro license), meaning not only the RAM is lower than in a premium capacity but also it is not dedicated to the workspace. 

2) Not sure that after 4 times of failed refresh Power BI disable it, but you can set an email notification when it happens.

[lbendlin · Accepted solution]

1) You're not the only workload on that capacity. Performance varies based on workload, time of day, seasonality etc

2) You need to distinguish between retries and full failures.  Your example doesn't count as it eventually succeeded. It has to completely fail all four tries  four times before the schedule is disabled.

Having said that - whatever you do to speed up the refresh is absolutely helping. Remember there is a hard limit (5 hours in Power BI service, but can be shorter in the data source) and you want to stay well under that.