[MDEV-27683] EXCHANGE PARTITION allows different index direction, but causes further errors Created: 2022-01-30  Updated: 2022-02-03  Resolved: 2022-02-03

Status: Closed
Project: MariaDB Server
Component/s: Partitioning, Server
Affects Version/s: 10.8
Fix Version/s: 10.8.1

Type: Bug Priority: Major
Reporter: Elena Stepanova Assignee: Sergei Golubchik
Resolution: Fixed Votes: 0
Labels: None

Issue Links:
Problem/Incident
is caused by MDEV-13756 Implement descending index: KEY (a DE... Closed

 Description   

EXCHANGE PARTITION allows a key to be ascending on one participating table and descending on another, but further DML on the table causes errors.

--source include/have_partition.inc
 
 
 
CREATE TABLE t1 (a INT, KEY(a DESC)) PARTITION BY KEY(a) PARTITIONS 4;
 
INSERT INTO t1 VALUES (6),(3),(2),(1);
 
 
 
CREATE TABLE t2 (a INT, KEY(a));
 
 
 
ALTER TABLE t1 EXCHANGE PARTITION p0 WITH TABLE t2;
 
INSERT INTO t1 VALUES (5),(4);
 
 
 
# Cleanup
 
DROP TABLE t1, t2;

10.8 0c5d1342 with MyISAM

 
mysqltest: At line 9: query 'INSERT INTO t1 VALUES (5),(4)' failed: ER_GET_ERRNO (1030): Got error 190 "Incompatible key or row definition between the MariaDB .frm file and the information in the storage engine. You may have retry " from storage engine MyISAM

With InnoDB INSERT doesn't fail, but errors are written in the log:

with InnoDB

 
ALTER TABLE t1 EXCHANGE PARTITION p0 WITH TABLE t2;
 
INSERT INTO t1 VALUES (5),(4);
 
DROP TABLE t1, t2;
 
bug.part2                                [ fail ]  Found warnings/errors in server log file!
 
        Test ended at 2022-01-30 15:14:13
 
line
 
2022-01-30 15:14:13 4 [ERROR] Found index a whose column info does not match that of MariaDB.
 
2022-01-30 15:14:13 4 [ERROR] InnoDB indexes are inconsistent with what defined in .frm for table ./test/t1#P#p0
 
^ Found warnings in /mnt8t/bld/10.8-debug-nightly/mysql-test/var/log/mysqld.1.err

With Aria, no errors are returned, which is more suspicious than soothing.

In MariaDB KB the description of EXCHANGE PARTITION is scarce, but MySQL manual says

Table nt is not itself partitioned.
Table nt is not a temporary table.
The structures of tables pt and nt are otherwise identical.
...

(there are 9 more points, but none of them changes the above)

However MySQL as of 8.0.28 also allows such EXCHANGE, and InnoDB errors/warnings also end up in the error log.

 Comments   
Comment by Elena Stepanova [ 2022-01-30 ]

Here is a slightly longer test case based on the same idea which ends with a server crash:

--source include/have_partition.inc
 
 
 
CREATE TABLE t1 (a INT, b INT, KEY (b DESC)) PARTITION BY KEY (b) PARTITIONS 4;
 
INSERT INTO t1 VALUES (8,6),(6,3),(9,2),(5,0);
 
 
 
CREATE TABLE t2 (a INT, b INT, KEY (b));
 
 
 
ALTER TABLE t1 EXCHANGE PARTITION p0 WITH TABLE t2;
 
SELECT * FROM t1 PARTITION (p1,p2);
 
 
 
--error ER_GET_ERRNO
 
INSERT INTO t1 VALUES (3,5);
 
DELETE FROM t1 PARTITION (p1,p2) WHERE a = 1 ORDER BY b, a LIMIT 1;
 
 
 
# Cleanup
 
DROP TABLE t1, t2;

10.8 0c5d1342

 
#3  <signal handler called>
 
#4  ha_partition::init_record_priority_queue (this=0x7faddc0f92f0) at /data/src/10.8/sql/ha_partition.cc:5450
 
#5  0x000055b7a031fec1 in ha_partition::index_init (this=0x7faddc0f92f0, inx=0, sorted=true) at /data/src/10.8/sql/ha_partition.cc:5587
 
#6  0x000055b79fb2c6d6 in handler::ha_index_init (this=0x7faddc0f92f0, idx=0, sorted=true) at /data/src/10.8/sql/handler.h:3446
 
#7  0x000055b79fb440b8 in init_read_record_idx (info=0x7fadedf83630, thd=0x7faddc000db8, table=0x7faddc06a708, print_error=true, idx=0, reverse=true) at /data/src/10.8/sql/records.cc:85
 
#8  0x000055b79fc13d63 in mysql_delete (thd=0x7faddc000db8, table_list=0x7faddc0155c8, conds=0x7faddc015e50, order_list=0x7faddc005ca8, limit=1, options=0, result=0x0) at /data/src/10.8/sql/sql_delete.cc:694
 
#9  0x000055b79fc7b865 in mysql_execute_command (thd=0x7faddc000db8, is_called_from_prepared_stmt=false) at /data/src/10.8/sql/sql_parse.cc:4804
 
#10 0x000055b79fc861c1 in mysql_parse (thd=0x7faddc000db8, rawbuf=0x7faddc015410 "DELETE FROM t1 PARTITION (p1,p2) WHERE a = 1 ORDER BY b, a LIMIT 1", length=66, parser_state=0x7fadedf84500) at /data/src/10.8/sql/sql_parse.cc:8027
 
#11 0x000055b79fc72822 in dispatch_command (command=COM_QUERY, thd=0x7faddc000db8, packet=0x7faddc00b999 "DELETE FROM t1 PARTITION (p1,p2) WHERE a = 1 ORDER BY b, a LIMIT 1", packet_length=66, blocking=true) at /data/src/10.8/sql/sql_parse.cc:1894
 
#12 0x000055b79fc711dc in do_command (thd=0x7faddc000db8, blocking=true) at /data/src/10.8/sql/sql_parse.cc:1402
 
#13 0x000055b79fe433c0 in do_handle_one_connection (connect=0x55b7a3cafb38, put_in_cache=true) at /data/src/10.8/sql/sql_connect.cc:1418
 
#14 0x000055b79fe4305f in handle_one_connection (arg=0x55b7a3cb21b8) at /data/src/10.8/sql/sql_connect.cc:1312
 
#15 0x000055b7a033718e in pfs_spawn_thread (arg=0x55b7a3caf6a8) at /data/src/10.8/storage/perfschema/pfs.cc:2201
 
#16 0x00007fadf3a8bea7 in start_thread (arg=<optimized out>) at pthread_create.c:477
 
#17 0x00007fadf368adef in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:95

Generated at Thu Feb 08 09:54:47 UTC 2024 using Jira 8.20.16#820016-sha1:9d11dbea5f4be3d4cc21f03a88dd11d8c8687422.