Liquid clustering on partitioned delta table where new data is added by replacing the partition

[SusmithaBadam]

Hi There,

Currently we have the bronze, silver & gold delta tables partition by a string and yearmonth column, we maintain the entire historic data of 5 years and replace the partitions (latest 2 months) to add the new data.

We use overwrite mode to replace the effected partitions.

Can liquid clustering be applied on these tables? I want to take the advantage of liquid clustering but want to overwrite the effected months only.

[MariuszK]

Hi,

You will need to update the data. Liquid clustering doesn't use physical folders for partitions, which means that you can't just override files in a folder with a partition.

Liquid clastering has its own benefits such incremental optimizalization, or effective data organziataion in files, but to migrate to Liquid clastering you will need to recreate tables.

[SusmithaBadam]

To take advantage of liquid clustering do I have to change the entire design of table to capture new data? between, which is the preferred mode or approach?

[SusmithaBadam]

But by removing the portioning I have to change the approach of capturing new data, is there any preferred way of capturing new data to apply LC?  

[szymon_dybczak]

Hi @SusmithaBadam ,

Short answer is no. Liquid Clustering is not compatible with partitioning or ZORDER. So, if you want to use LC you need to first remove partitioning from your tables.

Use liquid clustering for tables - Azure Databricks | Microsoft Learn