convert_to_rowstore()

Move a chunk from the columnstore to the rowstore

Since 2.18.0

convert_to_rowstore() replaces decompress_chunk(), deprecated in 2.18.0.

Manually convert a specific in the to the .

If you need to modify or add a lot of data to a in the , best practice is to stop any jobs moving s to the , convert the back to the , then modify the data. After the update, convert the to the and restart the jobs. This workflow is especially useful if you need to backfill old data.

Samples

To modify or add a lot of data to a :

1. Stop the jobs that are automatically adding s to the

   Retrieve the list of jobs from the timescaledb_information.jobs view to find the job you need to alter_job.

   SELECT alter_job(JOB_ID, scheduled => false);
   

2. Convert a to update back to the

   CALL convert_to_rowstore('_timescaledb_internal._hyper_2_2_chunk');
   

3. Update the data in the you added to the

   Best practice is to structure your INSERT statement to include appropriate partition key values, such as the timestamp. adds the data to the correct :

   INSERT INTO metrics (time, value)
   VALUES ('2025-01-01T00:00:00', 42);
   

4. Convert the updated s back to the

   CALL convert_to_columnstore('_timescaledb_internal._hyper_1_2_chunk');
   

5. Restart the jobs that are automatically converting s to the

   SELECT alter_job(JOB_ID, scheduled => true);
   

Arguments

The syntax is:

CALL convert_to_rowstore(
    chunk = '<chunk_name>',
    if_columnstore = true | false
);

Name	Type	Default	Required	Description
chunk	REGCLASS	-	✔	Name of the to be moved to the .
if_columnstore	BOOLEAN	true	✖	Set to false so this call fails with an error rather than a warning if chunk is not in the

Returns

This function returns void.