@Rocky is there a way to know size of individual columns on disk for tablet or partition?

To determine the disk size of individual columns for a specific tablet or partition in StarRocks, you can use the built-in metadata inspection functions. 1. Using

column_compressed_size

and

[_META_]

Hint StarRocks provides specialized functions to inspect column-level metadata. These functions must be used with the

[_META_]

hint, which directs the query engine to perform a

META_SCAN

on segment file headers rather than scanning the actual data. *

column_compressed_size(column_name)

: Returns the actual on-disk (compressed) size in bytes. *

column_size(column_name)

: Returns the decompressed size in bytes. To get column sizes for a specific tablet: You can include the hidden column

__tablet_id

in your query to group the results.

sql
SELECT 
    __tablet_id,
    SUM(column_compressed_size(your_column_name)) AS column_disk_size
FROM your_table_name [_META_]
-- Optional: filter by specific tablet
-- WHERE __tablet_id = 12345 
GROUP BY __tablet_id;

To get column sizes for a specific partition: You can filter the query by your partition key or use the

PARTITION

keyword to narrow down the scope.

sql
SELECT 
    SUM(column_compressed_size(column1)) AS col1_size,
    SUM(column_compressed_size(column2)) AS col2_size
FROM your_table_name PARTITION (p202301) [_META_];

2. Using System Statistics Tables If you don't need real-time precision and want to check general storage distribution, you can query the internal statistics database. StarRocks periodically collects column-level data size metrics for the Cost-Based Optimizer (CBO).

sql
SELECT 
    partition_name, 
    column_name, 
    data_size -- Estimated size in bytes
FROM _statistics_.column_statistics
WHERE table_name = 'your_table_name';

Note: These values are estimates based on the last statistics collection task. 3. Summary of Granularity Options Requirement | Recommended Method Exact size on disk (Per Column) |

SELECT column_compressed_size(col) FROM table [_META_]

Size per Tablet (Total) |

SHOW TABLET FROM table

or

SELECT * FROM information_schema.be_tablets

Size per Partition (Total) |

SHOW PARTITIONS FROM table

or

information_schema.partitions_meta

Important Note: The

[_META_]

hint and

column_compressed_size

functions are available in StarRocks v3.1 and later. If you are on an older version, column-level storage information is primarily available via the

_statistics_

database. References • sql-reference/sql-functions/utility-functions/column_size_functions.md • sql-reference/information_schema/be_tablets.md • sql-reference/information_schema/columns.md • sql-reference/sql-statements/table_bucket_part_index/SHOW_TABLET.md