Does dataset refresh support rollback?

I had defined unsupported textqualifier then refreshed the dataset.
Of course I got error as it failed but my previous existing data got wiped out.

Don’t we have rollback feature for Dataset refresh to prevent losing data like this event ?

Hi @tbdori - I do not think roll back option is available in SPICE data refresh. When you refresh the data, it wiped out the data and load new data, kind of cache solution.

Hi @Ashok @Max - Please correct me or advise on this.

Regards - Sanjeeb

When you’re editing your dataset, go to Settings > Publishing History.

You can then revert to an older version.


Hi, David

I was able to revert to a previous version.
So, I have v1 10 records, v2 12 records. Then I reverted to v1.
After save-publish, however, I still see 12 records in the dataset.

What are we reverting here? Just data schema like Physical/LogicalTable mapping only? The file itself contains 12 records at the moment.

My scenario, I had a version 1 with all columns,
and had version 2 with less columns. i.e I deleted 2 columns form the file.

In this case, I can’t go back to version1 as version 1 expect those 2 columns that I deleted from the file. So, it looks like that revert feature is there but supporting data must be in-sync with the version.


Hi @tbdori,

This just rolls back changes that you performed at the dataset level. It sounds like the changes were in your data source. You can’t undo those changes in QuickSight. QuickSight doesn’t keep the previous versions of your file.


Thank you @David_Wong . So if a data set is in SPICE and there is no change in the layout, we can revert back to the older version. Good to know that. Thanks for correcting my assumption as I thought of it worked like cache option but it is beyond that… great feature infact.

Regards - Sanjeeb

This option seems to be available even in datasets that use direct query. I think QuickSight just saves the dataset definition every time you publish. You can revert to an older definition but it may no longer be compatible with your source if the source has changed since (which is what happened here).


Thanks @David_Wong . It make sense when the data structure changed, the earlier layout seems to be invalid.

Thanks for sharing quick tips as always.

Regards - Sanjeeb