Lakehouse Recovery or Restore Points

[dwilliams3]

Lakehouses, and more importantly the data inside them, are not recoverable if the the lakehouse is deleted. There is also no way to recover a prior lakehouse if the data needs to be rolled back to a previous state.

BCDR was recently released but is only for capacity disaster recovery and is too cumbersome as it has an extra cost along with deploying in another region.

We need a way to recover a lakehouse and its history, if it has been deleted or if the data has been corrupted.

[tvilutis1]

We are evaluating this idea and its timeline along with other priorities. Your comments and votes will help us plan and design the feature. Thank you for your feedback! 

[daan2]

Is there any update on this feature request?

To me, this is a core feature that a production ready system must have.

This is currently keeping our company from using Fabric when building solutions.

[fbcideas_migusr]

I have voted for this idea.

My thoughts at first glance:

I think when someone deletes a Lakehouse, then the Lakehouse (including it's data) should be moved to a "recycle bin" for a retention period (e.g. 30 days) before it gets completely deleted.

In this period when the deleted lakehouse is in the "recycle bin", it should be possible to recover the Lakehouse (including it's data) by using an easy out-of-the-box solution.

The same should also be possible at Lakehouse table level.

Option to adjust settings in admin portal:

Preferably, someone like the capacity admin can adjust the retention period in order to balance storage costs vs. recovery possibility time window.

Also, someone like the capacity admin should have the option to allow selected users (security group) or all users to delete items from the "recycle bin", in this case the item gets completely deleted and unrecoverable. In order to save storage costs if necessary or delete unwanted data.

[yong_shao]

In production, this feature is very import to restore the lakehouse at ETL starting point when ETL process fails.

Snowflake has one command to restore its warehouse to a time point

[mse3]

This is what keeps us from using Fabric - customer wants a backup of lakehouse and also the retention period should be more than 30 days

[jtully]

This feature is a huge miss on the MS side of things. It's a very common user experience in other competitive products, and needs to be something that's available soon, especially since Fabric doesn't allow for explicit user permissions to prevent Contributors from accidentally deleting something like a lakehouse on accident. We can't hamstring our users into just using view only permissions, in fear of them deleting something for good.

[KimTutein]

A feature to recover deleted data from Lakehouse is allso a major thing holding us back in moving to Fabric. I am doing POC on Fabric (moving from Synapse to lakehouse / warehouse) but will have to have a plan for recoverying lakehouse and warehosue data. That should include someone deleting the whole lakehouse (I am aware of the soft delete in lakehouse on single tables can be reversed using Azure Storage browser). Perhaps this new feature same feature should be able to backup files somewhere else - that is disastor recovery if hackers got to your data and you needed to spin up a whole new tenant and restore lakehouses and warehouse there.

[JosAsh]

Basic backup and data recovery features aren't 'nice to haves' for large publicly listed companies.. Just sayin'.

[HatemMahdy]

In our case, we lost a heavy development on SQL server. So, only db schema backup was enough, because the data itself was synced from d365 fno and Dataverse. But, unfortunately, we lost even the schema. So, it is important feature to have, and schema backup doesn't cost big storage.