can anyone tell me if I did anything wrong in my design? Because it seems for some reasons, the memory used for processing tabular model is not released correctly.
The whole story is:
I created a SSIS package to loop all Azure SSAS tabular models, and then process them by pre-defined types, e.g. full, dataOnly + recalculate, clear + full, etc. If any error occurs, the package will log the error and process next model.
In my test, I processed 40+ models in full process mode, most of them are very small except one could be around 400mb. Some of them generated errors as I expected. But then I noticed that the memory usage of the SSAS instance increased dramatically.
I tried to run dmv queries to see sessions and sinkable memory allocations, but couldn't find any clues. At last I have to restart the instance to release the memory. As demonstrated, after restart of the service, memory usage dropped to just 5GB.
To my understanding when I selected full process mode, the memory used to process should be released. But in below demonstration, the memory was not released correctly.
Did anyone encounter this problem before? Is it related to the loop + full process? and if yes what is the correct way to process multiple tabular models?
Thanks in advance.
