# A way to replace the definition of a VDS? **URL:** \ **Category:** Uncategorized\ **Created:** [January 23, 2019, 9:47am UTC](https://community.dremio.com/t/a-way-to-replace-the-definition-of-a-vds/2604 "2019-01-23T09:47:31Z")\ **Posts on this page:** 7\ **Page:** 1
**Author:** ![dfleckinger](https://sea2.discourse-cdn.com/flex020/user_avatar/community.dremio.com/dfleckinger/32/370_2.png) [@dfleckinger](https://community.dremio.com/u/dfleckinger)\ **Post date:** [January 23, 2019, 9:47am UTC](https://community.dremio.com/t/a-way-to-replace-the-definition-of-a-vds/2604/1 "2019-01-23T09:47:31Z")
Hi, there is currently a way to create a VDS using command CREATE VDS, and there is a way to create a table through CREATE TABLE AS SELECT (CTAS). I would like to be able to replace a VDS and also replace a physical table ? Is there a command for that ? My use case is the following : I create a physical dataset using CTAS. I would like to periodically update this physical dataset (eg with new daily data) without loosing the accelerations already created. The data fields would remain the same. Thanks David ---
**Author:** ![kelly](https://sea2.discourse-cdn.com/flex020/user_avatar/community.dremio.com/kelly/32/11_2.png) [@kelly](https://community.dremio.com/u/kelly)\ **Post date:** [January 25, 2019, 5:33pm UTC](https://community.dremio.com/t/a-way-to-replace-the-definition-of-a-vds/2604/2 "2019-01-25T17:33:06Z")
What you might be able to do is to partition your data by time using CTAS: /mytable/day1/files.parquet /mytable/day2/files.parquet etc Then have a VDS that sits on top of /mytable. This would only work in an append-only model, no ability to update individual rows, but you could re-run the CTAS for a given day. Does that help? ---
**Author:** ![dfleckinger](https://sea2.discourse-cdn.com/flex020/user_avatar/community.dremio.com/dfleckinger/32/370_2.png) [@dfleckinger](https://community.dremio.com/u/dfleckinger)\ **Post date:** [January 29, 2019, 9:30am UTC](https://community.dremio.com/t/a-way-to-replace-the-definition-of-a-vds/2604/3 "2019-01-29T09:30:46Z")
Thanks @kelly. I’ve already implemented mechanisms such as the one you describe. From your answer, I understand there is currently no command to change the definition of a VDS, without removing it. I was hoping there was a command to mimic the “Save” button that we use when updating a dataset. ---
**Author:** ![pbofonseca](https://avatars.discourse-cdn.com/v4/letter/p/74df32/32.png) [@pbofonseca](https://community.dremio.com/u/pbofonseca)\ **Post date:** [August 19, 2019, 12:38pm UTC](https://community.dremio.com/t/a-way-to-replace-the-definition-of-a-vds/2604/4 "2019-08-19T12:38:19Z")
Hi, @dfleckinger Did you find any solution to refresh without recreate your dataset based on CTAS? I have something like that. I saved the results in a parquet file in S3, but I need to increment it. ---
**Author:** ![dfleckinger](https://sea2.discourse-cdn.com/flex020/user_avatar/community.dremio.com/dfleckinger/32/370_2.png) [@dfleckinger](https://community.dremio.com/u/dfleckinger)\ **Post date:** [August 19, 2019, 1:02pm UTC](https://community.dremio.com/t/a-way-to-replace-the-definition-of-a-vds/2604/5 "2019-08-19T13:02:40Z")
hi @pbofonseca, not really. I could not use Dremio for that, as during my tests CTAS was not able to create parquet files without also creating a Physical dataset at the same time. Also at the time, parquet files created by Dremio were not readable by my other python tools (pandas, pyarrow). I think that since a few versions, pyarrow pandas should be able to read Dremio CTAS files so may be there’s a way to use CTAS. What I currently do : - I have created a Dremio Physical Dataset (PDS) at the root. - For new data to be appended to this dataset, I extract data to a pandas DataFrame with Dremio, and write it as a parquet partitioned dataset on a daily basis within the PDS - After parquet writing, I execute the Dremio command “ALTER PDS {my\_pds} REFRESH METADATA”, so that Dremio is able to see the files that have just been created. ---
**Author:** ![dfleckinger](https://sea2.discourse-cdn.com/flex020/user_avatar/community.dremio.com/dfleckinger/32/370_2.png) [@dfleckinger](https://community.dremio.com/u/dfleckinger)\ **Post date:** [August 19, 2019, 1:07pm UTC](https://community.dremio.com/t/a-way-to-replace-the-definition-of-a-vds/2604/6 "2019-08-19T13:07:53Z")
Also I recently saw that there is now a way ro replace a VDS : [https://docs.dremio.com/sql-reference/sql-commands/datasets.html](https://docs.dremio.com/sql-reference/sql-commands/datasets.html) `> CREATE OR REPLACE VDS AS ` For example `> CREATE OR REPLACE VDS demo.jobs_vds as SELECT * FROM "oracle_e2e".DREMIO.JOBS` I don’t know which version introduced that but that’s a good new ! ---
**Author:** ![pbofonseca](https://avatars.discourse-cdn.com/v4/letter/p/74df32/32.png) [@pbofonseca](https://community.dremio.com/u/pbofonseca)\ **Post date:** [August 19, 2019, 2:19pm UTC](https://community.dremio.com/t/a-way-to-replace-the-definition-of-a-vds/2604/7 "2019-08-19T14:19:45Z")
> [@dfleckinger](#): > > - I have created a Dremio Physical Dataset (PDS) at the root. > - For new data to be appended to this dataset, I extract data to a pandas DataFrame with Dremio, and write it as a parquet partitioned dataset on a daily basis within the PDS > - After parquet writing, I execute the Dremio command “ALTER PDS {my\_pds} REFRESH METADATA”, so that Dremio is able to see the files that have just been created. I’ll try this! > [@dfleckinger](#): > > [https://docs.dremio.com/sql-reference/sql-commands/datasets.html](https://docs.dremio.com/sql-reference/sql-commands/datasets.html) I think that is my last step. Thanks a lot!