[Feature]: Drop a chunk by name

[lmandras-iura]

What problem does the new feature solve?

I need to deleted data based on the segment_by column (eg: device ID). Given i have a device ID, delete all chunks (including compressed) that belong to that device ID.

The point being to not use a regular DELETE sql statement that will have to decompress those chunks first and then run the delete. For GDPR purposes if i want to erase all data related to a device ID i want to do it as fast and simple as possible.

What does the feature do?

Drop/delete a chunk (including compressed) based on the chunk name (there is a query to get all chunks for a certain segment by value)

Implementation challenges

No response

[svenklemm]

With 2.17 you will be able to delete along segmentby columns without decompression.
This was added in #6882

[svenklemm]

As a workaround you can run manual DROP TABLE on those chunks. You can get a list of chunks matching arbitrary constraints like so:

                tableoid                
----------------------------------------
 _timescaledb_internal._hyper_1_2_chunk
(1 row)

[brasic]

It's not documented as far as I can see but _timescaledb_functions.drop_chunk() works for my use case. It takes a schema-qualified chunk name like

SELECT _timescaledb_functions.drop_chunk('_timescaledb_internal _hyper_3_230_chunk')

[svenklemm]

You can also do DROP TABLE _timescaledb_internal _hyper_3_230_chunk;