get_query_dump_from_query_id

get_query_dump_from_query_id(query_id) get_query_dump_from_query_id(query_id, enable_mock)

Returns a dump of a previously executed query, located by its query_id. The dump contains the table schemas, statistics, session variables and other context needed to reproduce planning of the original query, in the same JSON shape produced by get_query_dump and the /api/query_dump HTTP endpoint.

This function is intended for debugging and post-mortem analysis. It looks up the query in StarRocks' in-memory query detail queue on the current FE, recovers the original SQL plus its catalog and database, and re-runs the dumper against that SQL.

Arguments

query_id: The query identifier, in the standard StarRocks UUID format (xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx). Type: VARCHAR.

enable_mock: (Optional) When TRUE, the dump uses mocked table / column names instead of the originals (handy for sharing the dump without leaking schema). Defaults to FALSE. Type: BOOLEAN.

Return Value

VARCHAR. JSON-encoded query dump. Throws an error if the query cannot be located, the recorded SQL was desensitized, or the caller is not allowed to read the query.

Operational requirements

The query detail queue is the source of truth, so the same constraints that apply to that queue apply here. The function validates the two FE-config gates upfront and returns a clear error when either is mis-set:

• enable_collect_query_detail_info must be true (default false). When unset, the function refuses immediately with query detail collection is disabled. Set FE config enable_collect_query_detail_info=true ... instead of silently producing a "not found" error.
• enable_sql_desensitize_in_log must be false. When set to true the recorded SQL is rewritten to a digest form and cannot be re-dumped; the function refuses upfront with SQL desensitization is enabled ....

Additional runtime constraints, surfaced after the upfront checks pass:

• The query must have been executed on the same FE that the function is invoked on. Detail data is per-FE in-memory state, not synchronized.
• The detail row must still be in the cache window controlled by query_detail_cache_time_nanosecond (default 30 seconds). Older queries are evicted by a background sweep.
• As defense in depth, even with desensitization currently off, if the recorded SQL was captured while desensitization was on (so the detail row holds the placeholder), the function refuses with original sql not retained.

Privilege

The caller must either match the original executor's full account identity (user@host, e.g. 'alice'@'10.0.0.1') or hold system-level OPERATE privilege. Username-only matches are rejected, so two distinct accounts that share a username but differ on host cannot read each other's queries.

Examples

-- A user looking up one of their own queries (assuming
-- enable_collect_query_detail_info = true on the FE).
mysql> SELECT get_query_dump_from_query_id('a1b2c3d4-e5f6-7890-abcd-ef0123456789')\G

-- Same, but with mocked schema for safer sharing.
mysql> SELECT get_query_dump_from_query_id('a1b2c3d4-e5f6-7890-abcd-ef0123456789', TRUE)\G

-- Typical error when the query is not in the cache window.
mysql> SELECT get_query_dump_from_query_id('00000000-0000-0000-0000-000000000000');
ERROR 1064 (HY000): Getting analyzing error. Detail message: Invalid parameter
get_query_dump_from_query_id: query_id not found in query detail queue: ...

See also

• get_query_dump — dump a SQL string supplied directly by the caller, without going through the query detail queue.