We we not generating digest for

DDL

like

CREATE TABLE AS SELECT

? https://github.com/StarRocks/starrocks/blob/main/fe/fe-core/src/main/java/com/starrocks/qe/ConnectProcessor.java#L351

// Build Digest and queryFeMemory for SELECT/INSERT/UPDATE/DELETE
        if (ctx.getState().isQuery() || parsedStmt instanceof DmlStmt) {
            String digest = digestFromLeader;
            if (digest == null && (Config.enable_sql_digest || ctx.getSessionVariable().isEnableSQLDigest())) {
                digest = computeStatementDigest(parsedStmt);
            }
            if (digest != null) {
                ctx.getAuditEventBuilder().setDigest(digest);
            }
            long threadAllocatedMemory =
                    getThreadAllocatedBytes(Thread.currentThread().getId()) - ctx.getCurrentThreadAllocatedMemory();
            ctx.getAuditEventBuilder().setQueryFeMemory(threadAllocatedMemory);
        }

@Rocky @Michael Taranov @Rotem Adhoh cc

Based on the code snippet and StarRocks' current implementation (as of v3.3.6 and later), SQL digests are indeed only generated for Query statements (like

SELECT

) and DML statements (like

INSERT

,

UPDATE

,

DELETE

). Why

CREATE TABLE AS SELECT

(CTAS) is not included: In the StarRocks FE codebase,

CreateTableAsSelectStmt

is categorized as a DDL (Data Definition Language) statement rather than a

DmlStmt

. The logic in

ConnectProcessor.java

specifically filters for: 1. `ctx.getState().isQuery()`: Typically true for

SELECT

and certain

SHOW

statements. 2. `parsedStmt instanceof DmlStmt`: Covers

INSERT

,

UPDATE

, and

DELETE

. Since CTAS is technically a "Create Table" operation that happens to contain a "Select" query, it is treated as a DDL schema change. Most DDLs (

ALTER

,

DROP

,

CREATE

) are excluded from digest generation because: * Frequency: DDLs are generally low-frequency operations compared to queries and DML. * Purpose: The SQL Digest feature was primarily designed to identify patterns in high-concurrency workloads to analyze query performance and resource consumption. Impact on CTAS Even though the "Select" portion of a CTAS can be resource-intensive and follow a repeated pattern (common in ETL jobs), the outer statement remains a DDL. Consequently: * CTAS operations will appear in the audit log, but the

digest

field will be empty. * You cannot cluster CTAS operations by their query structure using the

digest

column in the

starrocks_audit_tbl__

. Recommendation If you need to track the performance patterns of heavy CTAS operations via digests, this would currently require an enhancement to the FE logic to include

CreateTableAsSelectStmt

(or DDLs in general) in the digest computation block. For now, if you are performing heavy ETL, you might consider using

INSERT INTO ... SELECT

into an existing table if possible, as

INSERT

statements will generate a digest. References: * SQL Digest Documentation * Audit Log Field Reference References • administration/sql_digest.md • administration/management/FE_configuration.md • faq/operation_maintenance_faq.md • administration/management/logs.md • administration/management/audit_loader.md • integrations/smt/starrocks_sqlalchemy.md