Closed
Description
Problem description
when used with pathman, SQL select from distribute table used Distributed Subplan 17_1
will report error.
CREATE TABLE t1 (id int, val int);
CREATE TABLE t2 (id int, val int);
SELECT create_distributed_table('t1', 'id');
SELECT create_distributed_table('t2', 'id');
xxx@postgres=# select* from t1,(select* from t2 where id>10) as tt;
ERROR: relation "17_1" does not exist
CONTEXT: while executing command on 192.168.247.128:6432
it is because citus will send 'copy xxx FROM STDIN WITH (format result)' to datanode and pathman is loaded after citus. without pathman, citus will process 'copy from' specially, it will create xxx for copy. but with pathman, pathman will to get relid of xxx in is_pathman_related_copy function, because xxx is not exist, it will throw exception, then select failed.
stack:
RangeVarGetRelidExtended
is_pathman_related_copy
pathman_process_utility_hook
Is it ok to change is_pathman_related_copy func as follows ?
parent_relid = RangeVarGetRelid(copy_stmt->relation,
(copy_stmt->is_from ?
PATHMAN_COPY_WRITE_LOCK :
PATHMAN_COPY_READ_LOCK),
- false);
-
+ true);
+ if (!OidIsValid(parent_relid))
+ return false;
+
/* Check that relation is partitioned */
if (has_pathman_relation_info(parent_relid))
{
(END)
Environment
....
....
22080 | pg_pathman | 10 | 2200 | f | 1.5 | {22082,22093
} | {"",""}
(28 rows)
pg version 13.3
Metadata
Metadata
Assignees
Labels
No labels