@Rocky there should be a parameter (variable) in starrocks shared cluster which tells analyze job to trigger external column statistics update if update_time > for instance 7 days and which one is that

In StarRocks (especially in versions 3.3 and 3.4+ which improved external statistics), the parameter that controls the interval for triggering automatic statistics updates for large tables—including external tables—is

statistic_auto_collect_large_table_interval

. Additionally, for the newer query-triggered analyze feature (v3.4.0+), the corresponding parameter is

connector_table_query_trigger_analyze_large_table_interval

. Key Parameters *

statistic_auto_collect_large_table_interval

: * Default:

43200

(12 hours). * Description: This FE configuration item defines the interval for automatically collecting statistics for large tables. For Hive and Iceberg tables, StarRocks checks if data has changed. If an update is detected, it only triggers a new

ANALYZE

job if the time since the last statistics update exceeds this interval. * To set this to 7 days, you would use:

604800

(seconds). *

connector_table_query_trigger_analyze_large_table_interval

(v3.4+): * Default:

43200

(12 hours). * Description: Specifically for external catalogs (Hive, Iceberg, Hudi, Delta Lake), this parameter controls how often a query can trigger a background

ANALYZE

task for a large table's columns. *

statistic_auto_collect_small_table_rows

: * Default:

10,000,000

. * Description: Determines the threshold for what StarRocks considers a "large table." Tables with row counts above this value follow the "large table" interval. How it works 1. Change Detection: For external tables like Hive and Iceberg, StarRocks periodically (every

statistic_collect_interval_sec

, default 5 mins) checks for metadata changes (e.g.,

transient_lastDdlTime

in Hive). 2. Interval Guard: Even if data has changed, the system will not trigger a new analyze job unless

Last Update Time + Collection Interval < Current Time

. This prevents expensive

ANALYZE

tasks from running too frequently on large external datasets. 3. Manual Override: When creating an automatic collection task via

CREATE ANALYZE TABLE

, you can also specify the property

"statistic_auto_collect_interval" = "604800"

to enforce a 7-day update cycle for that specific table. Configuration You can check or set these via the following SQL commands:

sql
-- Check current configuration
ADMIN SHOW CONFIG LIKE '%interval%';

-- Update configuration (e.g., set to 7 days)
ADMIN SET FRONTEND CONFIG ("statistic_auto_collect_large_table_interval" = "604800");
ADMIN SET FRONTEND CONFIG ("connector_table_query_trigger_analyze_large_table_interval" = "604800");

If you are looking for a parameter that defaults to 7 days, note that while

sys_log_delete_age

defaults to

7d

, the statistics intervals usually default to 12 hours. You likely want to modify the intervals mentioned above to achieve your 7-day requirement. References • using_starrocks/Cost_based_optimizer.md • sql-reference/information_schema/statistics.md • knowledge/usage/best-practice.md