Oracle Analytics Cloud and Server
Anyone else experience extremely slow response when creating a new dataset in DV?
Summary:
Sometimes when creating a new dataset in DV, the system seems to just hang when trying to profile the data or save the dataset. I can query the data directly and get results very quickly so I know this isn't a database issue. Is there a known cause? Is there a way to fix this? My system was OK this morning but now is just spinning.
Content (required):
Version (include the version you are using, if applicable):
September, 2022. OAC in a private subnet in OCI, connecting to an Oracle Autonomous Database in a private subnet in OCI.
Code Snippet (add any code snippets that support your topic, if applicable):
Answers
-
Hi Jerry,
Is it a particular that is taking time or in general even a simple select ?
Regards,
Narayanan
-
Hi Jerry,
For ADW I have checked in the instance. For ex- If I use a table with around 9000 rows of data and 35 columns, it takes around 20 seconds to load it in DV dataset.
I also assume the default analysis(shown below) DV does for each column might take some time. Not sure if this is something we can disable to save time.
-
Hi Jerry ,
If its taking time for profiling then it depends on the data volume of the underlying data set . Also there is a profile recommendations , which provides recommendations based on the number of columns in the data set .For varchar and date columns it recommends lot of profiling options . I think we can turn off the recommendations and then check if the performance is better or not .
Thanks
Anirban
-
Hi @Anirban Basu-133546, Any idea how to disable/turn off the profiling/recommedations option?
-
Hi Jahnavi ,
One option we can try is to manually delete the recommendations and then save the data set . From next time onwards , if we open the same data set then those recommendations won't come up.
Thanks
Anirban
-
In the September release, we added an option to "go right to editing" the metadata. The profiling and recommendations will still happen but move to the background. Many sources are fast and this is not necessary but some sources or tables take longer to profile and this can help.
-
This is not a consistent issue. Today I was able to create a dataset in under a minute that took a very long time last week. It seems to vary a different times during the day and in prod or non prod. It's inconsistent.
-
The video of the the September update profiling improvement features is here:
https://www.youtube.com/playlist?list=PL6gBNP-Fr8KUxQGQ8fKkv8dvQhb_mzygH
