2

I have a large table with roughly 90 million rows and 32 columns, all other tables process fine except for this one. I receive the following error.

Failed to save modifications to the server. Error returned: 'The command has been canceled.. The exception was raised by the IDataReader interface.

Technical Details: RootActivityId: b40bd1af-bf8b-40ea-acb7-748ae1bc7ae9 Date (UTC): 12/27/2017 2:50:52 PM '.

{ "refresh": { "type": "automatic", "objects": [ { "database": "DBName", "table": "TableName", "partition": "2017112" } ] } }

I have attempted to partition the table into smaller logical chunks, but still receive the same error, I suspect it is a size issue and going to continue attempting smaller partitions.

My source is Azure DW in to Azure SSAS.

Has anyone encountered this error? I tried googling to no avail? Is there a place I can lookup error meanings? Am I hitting memory constraints? How can I find out?

I know I can model the data better aligning with best practice, but due to business and resource constraints thats not a simple option (long story).

Im using Pricing tier D1

Any input welcome!

Thanks,

Ahmad B
  • 21
  • 1
  • 2

1 Answers1

0

I'm getting this issue on the D1 pricing tier which currently has up to 3GB RAM limit. On the Azure Analysis Service dashboard in Monitoring > Metrics > Memory I can see memory usage spike to just under 3GB before I get this error. This indicates that it's due to not enough memory.

When I change my model to retrieve less rows so that memory usage is less than 3GB I don't see the error and my model is deployed as expected.

While it's difficult to estimate the memory footprint in SSAS tabular models I would expect 90M rows and 32 columns to go well over the memory limit of the D1 tier. At about 10M rows with a similar number of columns I'm right on the limit of 3GB.

Peadar Doyle
  • 1,064
  • 2
  • 10
  • 26
  • 1
    Thanks for feedback, I have come to the same conclusion as well when I monitored the metrics. this was quite a disappointing limitation, as the PowerBi solution initially created held all the data(nearly reaching the limit) while my proof of concept in AAS worked but could not hold enough data to present its viability to the business. – Ahmad B Jan 24 '18 at 12:07